Mine roof dynamic disaster early warning method and system based on multi-point displacement data

Abstract

The application provides a mine roof dynamic disaster early warning method and system based on multi-point displacement data. The method firstly carries out superposition processing on the collected multi-point displacement monitoring data, and then establishes an autoregressive moving average model (ARIMA) model based on historical data matching and weighting, to realize accurate prediction of the superposed displacement data. Then, the characteristic function values of the superposed displacement data and the predicted data are calculated, i.e. the windowed waveform characteristic function ratio and the kurtosis value, and the mine roof dynamic disaster grading early warning is carried out, to issue green early warning, yellow early warning and red early warning. In summary, the technical scheme of the application enhances the signal-to-noise ratio of the displacement time series data using superposition technology, and solves the problem of limited adaptability of the absolute early warning threshold by using the windowed waveform characteristic function ratio. The linkage of the monitoring data and the predicted data realizes early and advanced early warning of the mine roof dynamic disaster, and the early warning result is more reliable and accurate than the single time series early warning method.

Description

Technical Field

[0001] This invention belongs to the field of engineering dynamic disaster monitoring technology, and in particular to a method and system for early warning of dynamic disasters on the top of a mine slab based on multi-point displacement data. Background Technology

[0002] During mining operations, the roof deforms due to stress. If this deformation exceeds its limit, localized roof falls or even large-scale collapses can occur, posing a significant challenge to safe mining operations. To address this, many mines have adopted support systems such as anchor mesh and anchor cables with metal mesh. However, due to factors such as fractured rock masses and invisible small faults, roof collapses still occur frequently. Accurate early warning systems for dynamic hazards on mine roofs will greatly reduce the risk of such disasters, ensuring the safety of workers and saving substantial time and economic costs.

[0003] Common analytical methods for mine dynamic hazards include theoretical calculations, similar physical simulation experiments, numerical simulations, and on-site monitoring. Mining is a complex system engineering project; theoretical calculations are often difficult to apply to complex structures, and similar physical simulation experiments and numerical simulations still have significant differences from actual mining conditions. On-site monitoring best reflects the real situation of mining operations. Currently, commonly used monitoring methods for mine dynamic hazards mainly include video surveillance, microseismic monitoring, ground sound monitoring, electromagnetic wave detection, stress monitoring, laser ranging, delamination monitoring, and displacement gauge monitoring. Among these, displacement monitoring methods such as laser ranging, delamination monitoring, and displacement gauges are characterized by their ease of monitoring and abundant data.

[0004] Traditional methods, based on displacement monitoring data, use displacement or acceleration thresholds to determine whether to issue an early warning. However, the early warning thresholds are usually different for different projects, which limits the application of such methods. Wang Dong et al. (2012) argued that the displacement velocity during the stable period of a slope follows a normal distribution, while it does not follow a normal distribution during the accelerated deformation stage. Therefore, they proposed a dynamic identification method for the imminent landslide moment based on the normality hypothesis testing theory. However, this method has high requirements for the normality of the displacement time series during the stable period, which can easily lead to false warnings. Wang Zhen et al. (2016), Cao Lanzhu et al. (2018), and Wang Dong et al. (2020) proposed Markov chain landslide early warning models based on mean-standard deviation, weighted average, and hierarchical clustering, respectively. These methods attempt to describe the slope evolution process from a probabilistic perspective. However, the Markov chain model prediction only considers the influence of the first few data points, ignoring the role of massive historical data. Wang Dong et al. (2022) proposed a method for slip surface identification and landslide early warning based on surface and deep displacement monitoring. This method uses the assumption that deep displacement precedes surface displacement to achieve graded early warning. However, unlike slope surfaces which are typically soft soil, mine roof slabs are usually hard rock, making the assumption of deep displacement preceding surface displacement difficult to apply. The existing patent, "A Deformation Monitoring Method for Collapse and Landslide Hazards," uses image comparison to determine the magnitude of displacement, and the ARIMA model used is a conventional model, which is insufficient to meet application requirements.

[0005] Therefore, this invention is dedicated to exploring how to reduce the false alarm rate of early warnings, make full use of historical data, and better achieve advanced early warning. Summary of the Invention

[0006] The purpose of this invention is to improve the accuracy of early warning for dynamic disasters on the top of mine roofs, and to provide a method and system for early warning of dynamic disasters on the top of mine roofs based on multi-point displacement data. The method provided by this invention acquires superimposed displacement time-series data with displacements at different depths through a deployed multi-point displacement monitoring system, achieving effective fusion of multi-source data. Secondly, it introduces an ARIMA model based on historical data matching to obtain predicted data, and utilizes both current and predicted data to achieve dual early warning, improving early warning accuracy and enabling advanced early warning of disasters.

[0007] On the one hand, the present invention provides a method for early warning of dynamic disasters on the top of a mine slab based on multi-point displacement data, comprising the following steps:

[0008] Step 1: Deploy a multi-point displacement meter monitoring system on the top of the mine to be monitored to obtain displacement time series data at different depths;

[0009] Step 2: Preprocess the collected displacement time series data at different depths to obtain superimposed displacement time series data. The preprocessing includes at least superposition processing, which is to superimpose the displacement time series data of multiple displacement gauges at the same time.

[0010] Step 3: Establish an ARIMA model based on historical data matching, and then predict the superimposed displacement time series data at future times based on the current superimposed displacement time series data and the ARIMA model;

[0011] Step 4: Calculate the feature values ​​of the current superimposed displacement time series data and the superimposed displacement time series data at future time points. The feature values ​​include at least the windowed waveform feature function ratio, which is calculated by comparing two window data of different lengths.

[0012] Step 5: Based on the aforementioned feature values, conduct graded early warning of dynamic disasters on the top of the mine roof.

[0013] Further, optionally, the ratio of the windowed waveform characteristic function in step 4 is expressed as:

[0014]

[0015] In the formula, The ratio of the characteristic function of the windowed waveform is denoted as x(i), which is the superimposed displacement value at time i in the current superimposed displacement time series data or the superimposed displacement time series data at future time; n and m are the lengths of the time series of windowing 1 and windowing 2, respectively, where m>n, such as n=3 and m=20.

[0016] The superposition process is expressed as follows:

[0017]

[0018] In the formula, This represents superimposed displacement time series data. , These are the displacement time series values ​​of the 1st and Nth points of the multi-point displacement gauge at time i, respectively.

[0019] Alternatively, step 3 can be implemented as follows:

[0020] Based on the autocorrelation function, identify the time series data in the historical superimposed displacement time series data that is most similar to the current superimposed displacement time series data;

[0021] The difference order d, the number of autoregressive terms p, and the number of moving average terms q of the ARIMA model are calculated using the closest time series data and its adjacent time series data; and these are used as the model parameters of the current ARIMA model.

[0022] Predict future stacked displacement time series data using the current ARIMA model parameters and current stacked displacement time series data.

[0023] Further, optionally, the number of autoregressive terms p and the number of moving average terms q in the ARIMA model are determined using a grid search weighted method, as follows:

[0024] First, the number of autoregressive terms p1 and the number of moving average terms q1 are determined using the autocorrelation coefficient (ACF) and the PACF index. Then, the range of values ​​for the number of autoregressive terms p and the number of moving average terms q are determined: p∈[p1-2, p1+2], q∈[q1-2, q1+2], and p, q≥0.

[0025] Then, within the range of values ​​for the number of autoregressive terms p and the number of moving average terms q, K ARIMA models are constructed by taking values ​​for the number of autoregressive terms p and the number of moving average terms q in turn, and the prediction performance of each ARIMA model is calculated separately.

[0026] Wherein, the prediction effect is the cumulative error between the predicted values ​​of the ARIMA model and the existing observations, and L is the length of the model. These are the predicted values ​​from the ARIMA model. For the observed values, Let be the cumulative error of the k-th ARIMA model;

[0027] Secondly, the weights of each ARIMA model are determined based on the cumulative error corresponding to each ARIMA model, where c is a preset coefficient. The weights are those of the k-th ARIMA model.

[0028] Finally, the final predicted value expression is obtained using the K ARIMA models, and the final number of autoregressive terms and moving average terms are obtained through back-inference, and assigned to the current ARIMA model; or the final predicted value expression obtained using the K ARIMA models is used as the prediction model for the current ARIMA model, and the predicted value is... Represented as: , These are the predicted values ​​of the 1st, 2nd, and Kth ARIMA models; These are the weights for the 1st, 2nd, and Kth ARIMA models.

[0029] Optionally, the feature value also includes kurtosis. Step 5 is to conduct a joint graded early warning of dynamic disasters on the top of the mine roof based on the ratio of the feature function of the windowed waveform and the kurtosis value. In this case, the feature function ratios of the current superimposed displacement time series data and the superimposed displacement time series data at future times are used to issue early warnings independently.

[0030] Specifically, two warning thresholds, R1 and R2, are set for the ratio of the characteristic function of the windowed waveform; and two warning thresholds, Ku1 and Ku2, are set for the kurtosis value Ku. The warning rules based on the ratio of the characteristic function of the windowed waveform R and the kurtosis value Ku are as follows:

[0031] If R < R1 and Ku < Ku1, it is regarded as safe;

[0032] If R1 ≤ R < R2 and Ku < Ku1, or if R < R1 and Ku1 ≤ Ku < Ku2, it is regarded as secondary;

[0033] In other cases, it is regarded as a first-level warning, and the first-level warning is more dangerous than the second-level warning.

[0034] Among them, the calculation method of the kurtosis value Ku is:

[0035]

[0036] In the formula: M is the set sliding time window length, is the average value of the sliding time window amplitude, is the standard deviation of the sliding time window amplitude, is the superimposed displacement.

[0037] Further optionally, when performing the dynamic disaster grading warning of the mine roof based on the ratio of the characteristic functions in step 5, the ratio of the characteristic functions of the current superimposed displacement time series data and the superimposed displacement time series data at future moments are independently warned:

[0038] Among them, first set two warning thresholds TH1 and TH2 that increase in sequence, and the warning rule based on the ratio R of the windowed waveform characteristic function is:

[0039] If R ≤ TH1, it is regarded as safe;

[0040] If TH1 ≤ R < TH2, it is regarded as a secondary warning;

[0041] If TH2 ≤ R, it is regarded as a first-level warning, and the first-level warning is more dangerous than the second-level warning.

[0042] On the other hand, the present invention provides a warning system based on the above method, including:

[0043] A data acquisition module, which is used to deploy a multi-point displacement meter monitoring system on the mine roof to be monitored to obtain displacement time series data at different depths;

[0044] A preprocessing module, which is used to preprocess the displacement time series data at different depths collected to obtain superimposed displacement time series data. The preprocessing at least includes superimposing processing, and the superimposing processing is: superimposing the displacement time series data of the multi-point displacement meters at the same moment;

[0045] A prediction module, which is used to establish an ARIMA model based on historical data matching, and then predict the superimposed displacement time series data at future moments based on the current superimposed displacement time series data and the ARIMA model;

[0046] The calculation module calculates the feature values ​​of the current superimposed displacement time series data and the superimposed displacement time series data at future times, respectively. The feature values ​​include at least the ratio of the windowed waveform feature functions, which is calculated by comparing two window data of different lengths.

[0047] The early warning module is used to perform graded early warning of dynamic disasters on the top of the mine roof based on the aforementioned feature values.

[0048] In three aspects, the present invention provides an electronic terminal, comprising at least:

[0049] One or more processors;

[0050] And a memory that stores one or more computer programs;

[0051] The processor invokes the computer program to achieve the following:

[0052] Steps of a method for early warning of dynamic disasters on the top of a mine roof based on multi-point displacement data.

[0053] In four aspects, the present invention provides a computer-readable storage medium storing a computer program, which is invoked by a processor for execution:

[0054] Steps of a method for early warning of dynamic disasters on the top of a mine roof based on multi-point displacement data.

[0055] Beneficial effects

[0056] 1. This invention provides a technical solution for early warning of dynamic disasters on mine roof slabs based on multi-point displacement data. This solution addresses the issue of early warning for mine roof slab disasters, enabling proactive disaster warnings, reducing engineering losses, and protecting personnel safety. First, the technical solution deploys a multi-point displacement monitoring system on the mine roof slab to acquire superimposed displacement time-series data with displacements at different depths, achieving effective fusion of multi-source data. This solves the problem of poor sensitivity and reliability of single-feature early warning in traditional multi-point displacement prediction methods due to the lack of multi-source data fusion. Second, the technical solution introduces an ARIMA model to predict future data, using current and predicted data for calculations to more comprehensively assess the current working conditions and improve the reliability of the early warning results. Furthermore, the technical solution sets a characteristic function ratio, which is closely related to the displacement change rate. The proposed characteristic function... Compared to the traditional amplitude value x(i), it can better capture the characteristic changes of the data, and thus reflect the displacement changes of the rock mass, achieving accurate early warning.

[0057] 2. The technical solution of this invention proposes a graded early warning system to achieve targeted early warning of dynamic disasters on the top of mine roofs, enabling the implementation of different measures. Specifically, considering the data characteristics of displacement time series data in this field, STA / LTA and kurtosis possess the ability to represent data statistical features, which can be used to characterize the relative changes in displacement data features. Furthermore, the characteristic function of STA / LTA has been improved, making the STA / LTA time series more recognizable. Therefore, STA / LTA and kurtosis are creatively selected for joint graded early warning. Compared with conventional absolute displacement / acceleration threshold early warning, the joint early warning of STA / LTA and kurtosis has a wider range of engineering applications and solves the problem of different early warning thresholds in different scenarios. Compared with the shortcomings of normality hypothesis testing, which has high data requirements and is prone to false alarms, STA / LTA and kurtosis do not have this assumption, resulting in a wider application range and a lower false alarm rate.

[0058] 3. The model parameters of the weighted ARIMA model proposed in this invention are not fixed. Instead, they are based on the current superimposed displacement time series data to match historical moments, obtaining the most relevant / similar time series data from the historical time series data. This allows for full utilization of historical data to determine the model parameters of the ARIMA model, improving the fit between the model parameters of the ARIMA model and the current working conditions, thereby improving the accuracy of the prediction results. At the same time, a grid search weighted algorithm is also selected to determine the current ARIMA model, improving the model prediction accuracy and reliability. Attached Figure Description

[0059] Figure 1 This is a flowchart of the method described in this invention;

[0060] Figure 2 This is a schematic diagram of the multi-point displacement gauge layout;

[0061] Figure 3 This is a graph of the original data of multi-point displacement;

[0062] Figure 4 This is a comparison chart of the original data and the filtered data;

[0063] Figure 5 This is a weighted diagram of K ARIMA models;

[0064] Figure 6 This is a comparison chart of the original data and the predicted data;

[0065] Figure 7 It is a comparison chart of superimposed displacement data;

[0066] Figure 8 It is a time series plot of the ratio of characteristic functions and kurtosis value of the windowed waveform. Detailed Implementation

[0067] The following will be combined with the appendix Figures 1-7 This paper further explains the method for early warning of dynamic disasters on the top of a mine slab based on multi-point displacement data proposed in this invention. Specific embodiments will be described in detail below.

[0068] Example 1:

[0069] like Figure 1 As shown, a method for early warning of dynamic disasters on the top of a mine roof based on multi-point displacement data includes the following steps:

[0070] Step 1: Deploy a multi-point displacement gauge monitoring system on the top of the mine to be monitored to obtain displacement time-series data at different depths. In this embodiment, a 3-point multi-point displacement gauge monitoring system is deployed on the top of the mine to obtain displacement time-series data at depths of 1 meter, 2 meters, and 3 meters. In other feasible embodiments, the model and type of the multi-point displacement gauge monitoring system are not restricted.

[0071] Step 2: Preprocess the collected displacement time series data at different depths to obtain superimposed displacement time series data. The preprocessing includes at least superposition processing.

[0072] In this embodiment, in addition to the overlay processing, moving average filtering and mean removal are also performed. The calculation formula for the overlay processing in this embodiment is as follows:

[0073] (1)

[0074] Where x(i) represents the superimposed displacement time series data. , , These are the displacement time series values ​​of the 1st, 2nd, and 3rd points of the multi-point displacement meter at time i.

[0075] It should be noted that moving average filtering and mean removal are existing technologies, and this invention does not elaborate on their implementation process. Furthermore, other feasible embodiments are not required to perform moving average filtering and mean removal. In this embodiment, the moving average filtering window length is 2 sampling points; the mean removal uses the average value of 30 points from the noise segment.

[0076] Step 3: Establish an ARIMA model (Autoregressive Integrated Moving Average Model) based on historical data matching, and then predict the superimposed displacement time series data for future times based on the current superimposed displacement time series data and the ARIMA model.

[0077] The ARIMA model is an existing model and its application in time series data prediction is a common application. Therefore, this invention does not elaborate on the specific construction and principles of this model. Unlike traditional methods that fix the model parameters of the ARIMA model, this embodiment uses dynamic ARIMA model parameters. The specific implementation process is as follows:

[0078] Based on the autocorrelation function, identify the time series data in the historical superimposed displacement time series data that is most similar to the current superimposed displacement time series data;

[0079] The difference order d, the number of autoregressive terms p, and the number of moving average terms q of the ARIMA model are calculated using the closest time series data and its adjacent time series data; and these are used as the model parameters of the current ARIMA model.

[0080] In this embodiment, a grid search weighted method is preferably used to determine the difference order d, the number of autoregressive terms p, and the number of moving average terms q of the historical data, as detailed below:

[0081] First, the initial number of autoregressive terms p1 and the number of moving average terms q1 are determined using the ACF and PACF indices. Then, the range of values ​​for the number of autoregressive terms p and the number of moving average terms q is determined: p∈[p1-2, p1+2], q∈[q1-2, q1+2], and p, q≥0.

[0082] Then, K ARIMA models are formed by taking values ​​sequentially within the range of values, and the prediction performance of each ARIMA model is calculated.

[0083] Wherein, the prediction effect is the cumulative error between the predicted values ​​of the ARIMA model and the existing observations, and L is the length of the model. These are the predicted values ​​from the ARIMA model. For the observed values, Let be the cumulative error of the k-th ARIMA model;

[0084] Secondly, the weights of each ARIMA model are determined based on the cumulative error corresponding to each ARIMA model, where c is a preset coefficient. The weights are those of the k-th ARIMA model.

[0085] like Figure 5 As shown, finally, the final predicted value expression is obtained using the K ARIMA models, and the final number of autoregressive terms and moving average terms are obtained through reverse reasoning and assigned to the current ARIMA model; or the final predicted value expression is obtained using the K ARIMA models and used as the prediction model of the current ARIMA model.

[0086] Predicted value Represented as: , These are the predicted values ​​of the 1st, 2nd, and Kth ARIMA models; These are the weights for the 1st, 2nd, and Kth ARIMA models.

[0087] In other feasible embodiments, the grid search method can be used to determine the difference order d, the number of autoregressive terms p, and the number of moving average terms q of the historical data, or other feasible methods can be used to obtain the difference order d, the number of autoregressive terms p, and the number of moving average terms q. The grid search method is briefly described as follows: determine the range of p to be 0~5, the range of q to be 0~5, and the range of d to be 0~2; traverse the values ​​within the range to construct various ARIMA models. The optimal p, q, and d values ​​are those that best match the prediction of the historical data with the subsequent sequence of the historical data sequence. For example, the final determined parameters are p=2, q=1, and d=1.

[0088] Step 4: Calculate the feature values ​​of the current superimposed displacement time series data and the superimposed displacement time series data at future time points.

[0089] In this embodiment, the preferred feature values ​​include the windowed waveform feature function ratio and the kurtosis value; in other feasible embodiments, only the windowed waveform feature function ratio may be set.

[0090] One type of characteristic function is defined as follows: The characteristic function set by the technical solution of this invention can better capture the characteristic changes of data compared with the traditional amplitude value x(i), thereby improving the reliability of the final warning result.

[0091] For the current stacked displacement time series data and the future stacked displacement time series data, calculate the time series of their respective windowed waveform characteristic function ratios, where the windowed waveform characteristic function is expressed as:

[0092] (1)

[0093] In the formula, The ratio of the characteristic function of the windowed waveform is given by x(i), where x(i) is the superimposed displacement value at time i in the current superimposed displacement time series data or the superimposed displacement time series data at future time; n is the length of the windowed 1 time series, n=3; and m is the length of the windowed 2 time series, m=20.

[0094] The kurtosis value Ku is calculated as follows:

[0095]

[0096] In the formula: M is the set sliding window length, This represents the average amplitude of the sliding window. is the standard deviation of the sliding time window amplitude, is the superimposed displacement.

[0097] Step 5: Based on the eigenvalue, perform classification and early warning for the dynamic disasters of the mine roof. The following will provide two embodiments to explain the classification and early warning:

[0098] (1) In this embodiment, two early warning thresholds R1 and R2 are set for the ratio of the windowed waveform characteristic function (STA / LTA); and two early warning thresholds Ku1 and Ku2 are set for the kurtosis value. The early warning rules based on the ratio R of the windowed waveform characteristic function and the kurtosis value Ku are as follows:

[0099] If R < R1 and Ku < Ku1, it is regarded as safe and a green light is displayed;

[0100] If R1 ≤ R < R2 and Ku < Ku1, or if R < R1 and Ku1 ≤ Ku < Ku2, both are regarded as level two and a yellow light is displayed;

[0101] In other cases, it is regarded as a level one warning and a red light is displayed. The level one warning is more dangerous than the level two warning.

[0102] It should be noted that the values of the early warning thresholds R1, R2 shown in Table 1 and the two early warning thresholds Ku1, Ku2 set for the kurtosis value are preferred values. In other feasible embodiments, adjustments can also be made according to the application scenario and the safety level. The specific values can be regarded as empirical values. As shown in Table 1 below:

[0103] Table 1 Joint early warning thresholds of STA / LTA and kurtosis

[0104]

[0105] (2) In other embodiments, if only the ratio of the windowed waveform characteristic function is considered as the eigenvalue, two gradually increasing early warning thresholds TH1 and TH2 are first set. The early warning rules based on the ratio R of the windowed waveform characteristic function are as follows:

[0106] If R ≤ TH1, it is regarded as safe;

[0107] If TH1 ≤ R < TH2, it is regarded as a level two warning;

[0108] If TH2 ≤ R, it is regarded as a level one warning. The level one warning is more dangerous than the level two warning.

[0109] If dynamic disaster grading and early warning for the mine roof are carried out, green prompts, yellow early warnings, and red early warnings are issued. Two early warning thresholds are set: TH1 = 4.0 and TH2 = 5.0. When R(i) ≤ TH1, it is safe and a green prompt is issued; when TH1 ≤ R(i) < TH2, a yellow early warning is issued; when TH2 ≤ R(i), a red early warning is issued. The yellow early warning indicates that the rock mass displacement accelerates, and preventive measures such as supporting the surrounding rock of the mining area and stress transfer are required; the red early warning indicates that the stability of the mine roof is very poor, and personnel must be immediately evacuated and relevant personnel notified.

[0110] It should be noted that the current superimposed displacement time series data and the superimposed displacement time series data at future moments are respectively early warned according to the above method, and each type of data will obtain a corresponding classification result. Both types of data have a warning effect. When the grading results of the two are different or the same, what specific measures to take or what warnings to issue are determined according to the specific working conditions and application requirements, and the present invention does not limit this. It should also be noted that the time series is continuous on the time axis. When early warning for a period of time series, even if different grading results appear at different time points within the same period, what specific measures to take or what warnings to issue are also determined according to the specific working conditions and actual application requirements, and the present invention does not limit this.

[0111] Step 6: Integrate the edge end of the mine roof early warning method to achieve real-time data processing and reduce equipment complexity.

[0112] Embodiment 2:

[0113] This embodiment provides an early warning system based on the above early warning method, which includes: a data acquisition module, a preprocessing module, a prediction module, an operation module, and an early warning module.

[0114] Among them, the data acquisition module is used to deploy a multi-point displacement meter monitoring system on the mine roof to be monitored to obtain displacement time series data at different depths; the preprocessing module is used to preprocess the acquired displacement time series data at different depths to obtain superimposed displacement time series data. The preprocessing at least includes superimposing processing, and the superimposing processing is: superimposing the displacement time series data of the multi-point displacement meter at the same moment; the prediction module is used to establish an ARIMA model based on historical data matching, and then predict the superimposed displacement time series data at future moments based on the current superimposed displacement time series data and the ARIMA model; the operation module is used to calculate the characteristic values of the current superimposed displacement time series data and the superimposed displacement time series data at future moments respectively; the early warning module is used to perform dynamic disaster grading and early warning for the mine roof based on the characteristic values.

[0115] It should be understood that the implementation process of each module can refer to the content description of the aforementioned method. The above division of functional modules is only a logical functional division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0116] Example 3:

[0117] This embodiment provides an electronic terminal, comprising at least: one or more processors; and a memory storing one or more computer programs; the processors invoke the computer programs to implement:

[0118] Steps of a method for early warning of dynamic disasters on the top of a mine roof based on multi-point displacement data.

[0119] For example, execute:

[0120] Step 1: Deploy a multi-point displacement meter monitoring system on the top of the mine to be monitored to obtain displacement time series data at different depths;

[0121] Step 2: Preprocess the collected displacement time series data at different depths to obtain superimposed displacement time series data. The preprocessing includes at least superposition processing, which is to superimpose the displacement time series data of multiple displacement gauges at the same time.

[0122] Step 3: Establish an ARIMA model based on historical data matching, and then predict the superimposed displacement time series data at future times based on the current superimposed displacement time series data and the ARIMA model;

[0123] Step 4: Calculate the feature values ​​of the current superimposed displacement time series data and the superimposed displacement time series data at future time points. The feature values ​​include at least the windowed waveform feature function ratio, which is calculated by comparing two window data of different lengths.

[0124] Step 5: Based on the aforementioned feature values, conduct graded early warning of dynamic disasters on the top of the mine roof.

[0125] The implementation process of each step can be referred to the detailed steps of Embodiment 1. The memory may include high-speed RAM memory, and may also include a non-volatile defibrillator, such as at least one disk storage device.

[0126] If the memory and processor are implemented independently, they can be interconnected via a bus to communicate with each other. This bus can be an industry-standard architecture bus, an external device interconnect bus, or an extended industry-standard architecture bus, etc. The bus can be categorized as an address bus, data bus, control bus, etc.

[0127] Optionally, in a specific implementation, if the memory and processor are integrated on a single chip, the memory and processor can communicate with each other through an internal interface.

[0128] It should be understood that, in the embodiments of the present invention, the processor may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor. The memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of the memory may also include non-volatile random access memory. For example, the memory may also store device type information.

[0129] Example 4:

[0130] This embodiment provides a computer-readable storage medium storing a computer program, which is invoked by a processor for execution.

[0131] Steps of a method for early warning of dynamic disasters on the top of a mine roof based on multi-point displacement data.

[0132] For example, execute:

[0133] Step 1: Deploy a multi-point displacement meter monitoring system on the top of the mine to be monitored to obtain displacement time series data at different depths;

[0134] Step 2: Preprocess the collected displacement time series data at different depths to obtain superimposed displacement time series data. The preprocessing includes at least superposition processing, which is to superimpose the displacement time series data of multiple displacement gauges at the same time.

[0135] Step 3: Establish an ARIMA model based on historical data matching, and then predict the superimposed displacement time series data at future times based on the current superimposed displacement time series data and the ARIMA model;

[0136] Step 4: Calculate the feature values ​​of the current superimposed displacement time series data and the superimposed displacement time series data at future time points. The feature values ​​include at least the windowed waveform feature function ratio, which is calculated by comparing two window data of different lengths.

[0137] Step 5: Based on the aforementioned feature values, conduct graded early warning of dynamic disasters on the top of the mine roof.

[0138] Please refer to the explanation of the method above for the specific implementation process of each step.

[0139] The readable storage medium is a computer-readable storage medium, which can be an internal storage unit of the controller described in any of the foregoing embodiments, such as the controller's hard drive or memory. The readable storage medium can also be an external storage device of the controller, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the controller. Further, the readable storage medium can include both the controller's internal storage unit and external storage devices. The readable storage medium is used to store the computer program and other programs and data required by the controller. The readable storage medium can also be used to temporarily store data that has been output or will be output.

[0140] Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned readable storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0141] Verification Instance

[0142] Figure 2 This is a schematic diagram of the multi-point displacement gauge deployment. Three multi-point displacement gauges are deployed on the top rock mass of the roadway to be monitored to monitor displacement at depths of 1 meter, 2 meters, and 3 meters. The multi-point displacement gauge data is collected, processed, and transmitted via an edge-end computing rod.

[0143] Figure 3 This is a graph of the original displacement data at multiple points. When the roof is in a stable period, the displacement data of each sensor approximately follows a Gaussian distribution; when the roof is in an unstable period, the sensor displacement data begins to increase. These characteristics provide a basis for early warning of dynamic disasters on the roof of the mine.

[0144] Figure 4This is a comparison chart of the original data and the filtered data. It can be seen that moving average filtering can reduce the oscillations in the displacement data. Furthermore, removing the mean can improve the signal-to-noise ratio of the displacement data and enhance the discriminative power of the windowed waveform characteristic function ratio.

[0145] Figure 5 This is a weighted diagram of K ARIMA models. The diagram illustrates the prediction process of the k-th ARIMA and the weighting principle.

[0146] Figure 6 This is a comparison chart of the original data and the predicted data. It shows that the ARIMA model based on historical data matching can accurately predict multi-point displacement data, providing a data foundation for early warning of dynamic disasters on the top of mine roofs.

[0147] Figure 7 This is a comparison chart of superimposed displacement data. It shows that the signal-to-noise ratio (SNR) of the superimposed original data is better than that of the original displacement data from each point on the multi-point displacement gauge, indicating that data superposition enhances the data features. This is because random data addition may cancel each other out, while feature data is enhanced. Furthermore, the SNR of the superimposed filtered data and the predicted data is better than that of the superimposed original data, confirming the significance of filtering.

[0148] Figure 8 This is a time series plot of the characteristic function ratio and kurtosis value of the windowed waveform. It can be seen that the characteristic function ratio of the windowed waveform at a single displacement point oscillates very largely, and the increase in the characteristic function ratio during the development of dynamic disasters on the mining roof is not significant. This is because, without filtering and mean removal processing, the increase in amplitude during the development of dynamic disasters is not significant compared to the original amplitude. The characteristic function ratio of the overlaid data increases, especially the characteristic function ratio after filtering and mean removal processing; the overlaid predicted data also shows a good characteristic function ratio. The kurtosis value shows a similar effect, further demonstrating the effectiveness of this patented method.

[0149] It should be emphasized that the examples described in this invention are illustrative rather than limiting. Therefore, this invention is not limited to the examples described in the specific embodiments. Any other embodiments derived by those skilled in the art based on the technical solutions of this invention, without departing from the spirit and scope of this invention, whether modifications or substitutions, are also within the protection scope of this invention.

Claims

1. A method for early warning of dynamic disasters on the top of a mine slab based on multi-point displacement data, characterized in that, It includes the following steps: Step 1: Install a multi-point displacement meter monitoring system on the roof of the mine to be monitored to obtain displacement time-series data at different depths; Step 2: Preprocess the collected displacement time-series data at different depths to obtain superimposed displacement time-series data. The preprocessing at least includes superimposing processing, and the superimposing processing is: superimposing the displacement time-series data of the multi-point displacement meter at the same moment; Step 3: Establish an ARIMA model based on historical data matching, and then predict the superimposed displacement time-series data at future moments based on the current superimposed displacement time-series data and the ARIMA model; Step 4: Calculate the eigenvalues of the current superimposed displacement time-series data and the superimposed displacement time-series data at future moments respectively. The eigenvalues at least include the ratio of windowed waveform feature functions, and the ratio of windowed waveform feature functions is calculated by taking the ratio of two window data with different lengths; Step 5: Conduct hierarchical early warning for dynamic disasters on the roof of the mine based on the eigenvalues; The implementation process of Step 3 is as follows: Based on the autocorrelation function, identify the time-series data in the historical superimposed displacement time-series data that is closest to the current superimposed displacement time-series data; use the closest time-series data and its adjacent time-series data to calculate the difference order d, the autoregressive term number p, and the moving average term number q of the ARIMA model; And use them as the model parameters of the current ARIMA model; use the model parameters of the current ARIMA model and the current superimposed displacement time-series data to predict the superimposed displacement time-series data at future moments; The autoregressive term number p and the moving average term number q of the ARIMA model are determined by the grid search weighting method as follows: First, use the ACF index and the PACF index to determine the preliminary autoregressive term number p1 and the moving average term number q1, and then determine the value ranges of the autoregressive term number p and the moving average term number q: p ∈ [p1 - 2, p1 + 2], q ∈ [q1 - 2, q1 + 2], and p, q ≥ 0; then, take values for the autoregressive term number p and the moving average term number q in turn within the value ranges of the autoregressive term number p and the moving average term number q to form K ARIMA models, and then calculate the prediction effects of each ARIMA model respectively; The prediction result is the cumulative error between the ARIMA model's predicted values ​​and existing observations. L is the length of the model. These are the predicted values ​​from the ARIMA model. Let be the observed value of the superimposed displacement at time i. Let be the cumulative error of the k-th ARIMA model; Secondly, the weights of each ARIMA model are determined based on the cumulative error corresponding to each ARIMA model. c is a preset coefficient; finally, the final predicted value expression is obtained using the K ARIMA models, and the final number of autoregressive terms and moving average terms are obtained through reverse reasoning, and assigned to the current ARIMA model; or the final predicted value expression is obtained using the K ARIMA models as the prediction model of the current ARIMA model, and the predicted value is... Represented as: , These are the predicted values ​​of the 1st, 2nd, and Kth ARIMA models; These are the weights for the 1st, 2nd, and Kth ARIMA models.

2. The method according to claim 1, characterized in that: The ratio of windowed waveform feature functions described in Step 4 is expressed as: ， In the formula, The ratio of the characteristic function of the windowed waveform is given, where n and m are the lengths of the time series of windowed waveform 1 and windowed waveform 2, respectively, and m > n.

3. The method according to claim 1, characterized in that: ​ ​ ​ ​ ​ 4. The method according to claim 3, characterized in that: ​ ， In the formula: M is the set sliding window length. This represents the average amplitude of the sliding window. This represents the standard deviation of the sliding window amplitude.

5. The method according to claim 1, characterized in that: When performing hierarchical early warning of dynamic disasters on the mine roof based on the eigenvalues in step 5, the current superimposed displacement time series data and the superimposed displacement time series data at future times are independently warned: Among them, first set two gradually increasing warning thresholds TH1 and TH2. The warning rule based on the ratio R of the windowed waveform characteristic function is: If R < TH1, it is regarded as safe; If TH1 ≤ R < TH2, it is regarded as a secondary warning; If TH2 ≤ R, it is regarded as a primary warning, and the primary warning is more dangerous than the secondary warning.

6. An early warning system based on the method of any one of claims 1-5, characterized in that, It includes: A data acquisition module, used to deploy a multi-point displacement meter monitoring system on the mine roof to be monitored to obtain displacement time series data at different depths; A preprocessing module, used to preprocess the displacement time series data at different depths collected to obtain superimposed displacement time series data. The preprocessing at least includes superimposing processing, and the superimposing processing is: superimposing the displacement time series data of the multi-point displacement meter at the same time; A prediction module, used to establish an ARIMA model based on historical data matching, and then predict the superimposed displacement time series data at future times based on the current superimposed displacement time series data and the ARIMA model; An operation module, which calculates the eigenvalues of the current superimposed displacement time series data and the superimposed displacement time series data at future times respectively. The eigenvalues at least include the ratio of the windowed waveform characteristic function, and the ratio of the windowed waveform characteristic function is to calculate the ratio of two window data with different lengths; An early warning module, used to perform hierarchical early warning of dynamic disasters on the mine roof based on the eigenvalues.

7. An electronic terminal, characterized in that: It at least includes: One or more processors; And a memory storing one or more computer programs; The processor calls the computer program to implement: The steps of the method according to any one of claims 1-5.

8. A computer-readable storage medium, characterized in that: Stored a computer program, and the computer program is called by the processor to execute: The steps of the method according to any one of claims 1-5.

Images