Stope roof significant weighting prediction method based on edge detection

Abstract

The invention discloses a stope roof significant weighting prediction method based on edge detection, and belongs to the technical field of coal mining, and the stope roof significant weighting prediction method comprises the following steps: step 1, obtaining real-time pressure data of a hydraulic support of an electro-hydraulic control system; step 2, preprocessing the pressure data, including abnormal value detection and processing, missing value processing and data denoising; 3, fitting the data points into a continuous curved surface through a curved surface fitting technology, projecting the continuous curved surface, and extracting spatial distribution characteristics of roof pressure; 4, calculating relevant parameters of the projection area; and step 5, performing significant weighting area prediction and parameter updating by using an edge detection technology. According to the method, the significant weighting area can be predicted more accurately, misjudgment and missed judgment conditions are reduced, and a more reliable basis is provided for coal mine safety production.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of coal mining, and particularly relates to a stope roof significant weighting prediction method based on edge detection. BACKGROUND

[0002] In the coal mining industry, the accuracy of roof weighting directly relates to safety production and mining efficiency. In the past, the industry generally used empirical formulas and simple statistical analysis methods to predict roof weighting. These traditional methods mainly rely on the experience of mining personnel, combined with limited hydraulic support pressure data, such as calculating the average, maximum and minimum values of pressure within a period of time, and then comparing them with the preset empirical threshold to infer the possibility of roof weighting.

[0003] However, the coal mining environment is extremely complex, and the mine pressure is influenced by a variety of factors. The overburden pressure is different at different mining depths; the pressure distribution pattern changes with the difference in ore body morphology, such as thickness and inclination; and the mining method, such as mining technology and mining sequence, also significantly affects the mine pressure. In such a complex situation, traditional empirical formulas and basic statistical analysis are difficult to fully and accurately reflect the change pattern of mine pressure. Taking the analysis of the three-dimensional spatial distribution of mine pressure as an example, due to the lack of deep mining ability of data, the traditional method cannot accurately locate the area of sharp pressure change, and it is also difficult to analyze the mutual relationship of pressure change in different areas, and cannot provide comprehensive and reliable data support for coal mining.

[0004] With the development of advanced technologies such as big data and machine learning, new ideas have been brought to the field of coal mining. Some research attempts to use machine learning algorithms to process mine pressure data, but the coal mining environment is special, the data collection is difficult and noisy, and the existing machine learning models are often too complex, requiring high computing resources, which is difficult to efficiently deploy and apply in the mine site. Moreover, these methods mostly do not fully combine the actual factors such as geological conditions and mining technology of coal mining, resulting in unsatisfactory accuracy and reliability of the prediction results.

[0005] In the process of coal mining, roof weighting is a key factor affecting safety production. The current traditional roof weighting prediction method relies on empirical formulas and simple statistical analysis, which is difficult to accurately predict the significant weighting area. The existing technology mainly has the following problems.

[0006] Data processing and analysis depth is lacking: the existing technology of mine pressure data processing mostly stays in simple statistics level, such as calculating average value, maximum value, minimum value, etc. Lack of effective data processing algorithm and model, can't fully tap the data characteristics and rules. Mine pressure is affected by depth, ore body shape, mining method and other factors, simple statistics can't reflect these complex relationships. For example, in the analysis of three-dimensional spatial distribution of mine pressure, it is difficult to accurately locate the pressure change area, and it is difficult to provide comprehensive and accurate data support for mine exploitation.

[0007] Two-dimensional analysis limitations: existing technology mostly uses two-dimensional analysis method, which cannot fully reflect the change rule of roof pressure in three-dimensional space. Roof movement is a complex three-dimensional process, and it is difficult to accurately capture its key features through two-dimensional analysis, resulting in insufficient accuracy and reliability of prediction results.

[0008] The prediction model lacks adaptability: most of the existing prediction methods are based on fixed data model and parameters, which cannot be dynamically adjusted with the progress of mining and the change of geological conditions. This makes the prediction model prone to prediction accuracy decline when facing complex and variable mining environment.

[0009] Based on this, the present application provides a kind of based on edge detection's stope roof significant pressure prediction method, by the analysis of historical data identifies significant area and calculates characteristic value, then further predicts the pressure area according to these results, to improve the accuracy and reliability of prediction, meet the needs of coal mining industry for safe and efficient mining. SUMMARY

[0010] In order to solve the above problems, the present application provides a kind of based on edge detection's stope roof significant pressure prediction method, through real-time acquisition and processing of electro-hydraulic control system hydraulic support pressure data, using surface fitting technology to construct three-dimensional surface, and analyzing significant pressure area by plane projection. At the same time, combined with edge detection technology to predict whether there will be significant pressure, identify the significant pressure area of mine and give early warning.

[0011] The technical scheme of the present application is as follows:

[0012] A kind of based on edge detection's stope roof significant pressure prediction method, characterized in that, including the following steps:

[0013] Step 1, obtain the real-time pressure data of electro-hydraulic control system hydraulic support;

[0014] Step 2, pre-process the pressure data, including outlier detection and processing, missing value processing and data denoising;

[0015] Step 3, fit the data points into continuous surface by surface fitting technology and project, extract the spatial distribution characteristics of roof pressure;

[0016] Step 4, calculate the projection area related parameters;

[0017] Step 5, use edge detection technology to perform significant area prediction and parameter update.

[0018] Further, in step 1, the pressure data of the hydraulic support is collected in real time by the pressure sensor equipped in the downhole electro-hydraulic control system according to the set high-frequency sampling frequency g; in a certain footage interval, the pressure value of the i-th support y i The pressure data sequence collected by the upper pressure sensor is denoted as z = [z i1 ,z i2 ,…,z ih ], where h is the number of sampling points in the footage interval, z ih represents the pressure value of the i-th support y h when the footage is x i ; finally, the collected pressure data is transmitted and stored in the database in real time through the data transmission network.

[0019] Further, the specific process of step 2 is as follows:

[0020] Step 2.1, outlier detection and processing;

[0021] For the pressure data sequence, the mean and standard deviation are calculated:

[0022]

[0023] Where μ i is the mean of the pressure data collected by the pressure sensor on the i-th support in a certain time period; z ij represents the pressure value of the i-th support y j when the footage is x i ; j is the index value of the footage; σ i is the standard deviation of the pressure data collected by the pressure sensor on the i-th support in a certain time period; N i is the total number of pressure data points collected by the pressure sensor on the i-th support;

[0024] If |z ij -μ i |>3σ i , then z ij is determined to be an outlier, which is replaced by the median M i of the sequence; M i is the median of all data values of the pressure sensor on the i-th support;

[0025] Step 2.2, missing value processing;

[0026] For data points with missing values, linear interpolation is used to fill them; the corresponding distance of the missing value is defined as x k , if z ik Missing, and z i(k+1) 、z i(k-1) Known, then:

[0027]

[0028] Among them, z ik 、z i(k+1) 、z i(k-1) Indicates the footage is x k 、x k+1 、x k-1 When the i-th bracket y i The pressure value on the y-axis; k, k+1, k-1 are index values ​​of different footages respectively;

[0029] Step 2.3: Data denoising;

[0030] The sliding average filter method is used to denoise the cleaned data; assuming the sliding window size is m, each pressure value is denoised, and the formula is:

[0031]

[0032] Among them, z′ ij Indicates footage is x j When the i-th bracket y i The pressure value after denoising; z il Indicates footage is x l When the i-th bracket y i The pressure value on the y-axis; l is the index value of footage;

[0033] When l<1 or l>N i When the forward filling method is used, that is, z il =z i1 or z i1 、 Indicates the footage is x1, When the i-th bracket y i The pressure value on.

[0034] Furthermore, the specific process of step 3 is as follows:

[0035] Step 3.1, fit the distribution of pressure data in three-dimensional space by surface fitting technology, and construct a top plate to press the surface;

[0036] Extract data points from the real-time pressure data of the hydraulic support preprocessed in step 2, construct a top plate in three-dimensional space to press the curved surface, set the advancing direction along the working surface as the x-axis, the vertical direction as the y-axis, and the pressure value as the z-axis; if the advance is x j When the i-th bracket y is collected i The pressure value is z ij , then it corresponds to the three-dimensional coordinate (x j ,y i ,z ij ), and thus construct the data point set of the top plate to press the surface:

[0037] {(x j ,y i ,z ij )|i=1,2,...,α;j=1,2,...,β};

[0038] Among them, α is the total number of brackets; β is the total number of footage indexes;

[0039] The top plate presses the surface Z = f(x, y), and the fitting model is:

[0040] Z=a0+a1x+a2y+a3x 2 +a4y 2 +a5xy;

[0041] Where f(x,y) is the pressure value collected by the pressure sensor of bracket y when the footage is x; a0, a1, a2, a3, a4, and a5 are different coefficients to be determined;

[0042] The error function E is constructed based on the top plate to compress the surface:

[0043]

[0044] Calculate the partial derivatives of the error function E of the top plate pressure surface with respect to the coefficients a0, a1, a2, a3, a4, and a5, and set the partial derivatives to 0, resulting in a system of six equations. Solve this system of equations to determine the values ​​of the coefficients a0, a1, a2, a3, a4, and a5, and then determine the specific expression for the top plate pressure surface Z = f(x, y).

[0045] Step 3.2, projection processing;

[0046] Set the pressure data value for cutting the surface to D0; use the plane Cut the determined top plate to press the curved surface, project the cut curve onto the cutting plane parallel to the xoy plane to obtain the projection area; at the same time, record the data point information in the projection area, including the pressure value, the corresponding bracket position and the acquisition time.

[0047] Further, the specific process of step 4 is as follows:

[0048] Step 4.1, calculate the area of the projection region and the data mean in the region;

[0049] Divide each projection region into MxN grids, each grid has a size of Δx x Δy, and count the number of grids falling within the projection region K, then the area S of the rth projection region is r

[0050] S r = K x (Δx x Δy);

[0051] Where Δx and Δy are the length and width of each grid respectively; M and N are the number of grids along the x-axis and y-axis respectively;

[0052] The total area S of the projection region is the sum of the areas of all projection regions:

[0053]

[0054] Where q is the number of projection regions, and r = 1, 2, …, q;

[0055] Calculate the area mean of all projection regions

[0056]

[0057] The set of pressure data points of the rth projection region is

[0058]

[0059] Where P rg is the value of the gth pressure data point in the rth projection region; g is the number of pressure data points in the rth projection region;

[0060] For the rth projection region, the pressure data mean

[0061]

[0062] Where P ro is the value of the oth pressure data point in the rth projection region;

[0063] Calculate the data mean of all projection regions

[0064]

[0065] Step 4.2, comparison and adjustment;

[0066] ​​​Calculate the total area S of the mined working face area total， The formula is:

[0067] S total =l push ×l ten ;

[0068] Among them, l push is the historical advancement distance of the working face; l ten is the inclined length of the working face;

[0069] Combine S with S total Compare and determine whether 0.15S is met total <S<0.2S total If the condition is not met, the value of D0 needs to be readjusted and the surface cutting and projection operations need to be performed again until the condition is met. The data analysis of the historical advancement distance is then completed and the mean value of the data calculated for all projection areas is recorded. and the mean area

[0070] Furthermore, the specific process of step 5 is as follows:

[0071] Step 5.1: Assume that the footage of the current forecast stage is Index value of footage in the prediction stage; connect the mine pressure data points in the current prediction stage into line segments The line contains multiple pressure data points For line segments On the pressure data points; is a line segment The total number of pressure data points on ;

[0072] Step 5.2: Use the sliding window method to traverse the line segment Determine the predicted significant pressure area segment and endpoint locations;

[0073] Step 5.3: Determine whether the current When satisfied When , perform the calculation in step 5.4; if Then calculate the next footage, that is, update the footage index value Switch to the next line segment Execute step 5.2 to continue sliding window traversal and condition judgment;

[0074] Step 5.4: Calculate the area ratio and set a warning threshold. When the area ratio is greater than the set warning threshold, trigger the warning mechanism and design a warning grading mechanism for graded warnings.

[0075] Step 5.5: If the footage If the distance exceeds 30m, the area and data of the significant pressure area of ​​the predicted part are added to the historical data, and the area mean and data mean of the significant pressure area are recalculated and updated, that is, an update is made. and If the advance If the footage index value does not exceed 30m, update the footage index value. Switch to the next line segment After the switch is completed, for the new line segment Repeat the sliding window analysis in step 5.2.

[0076] Furthermore, the specific process of step 5.2 is as follows:

[0077] Step 5.2.1. Set a fixed-length sliding window with a window length of w and a sliding window step size of S. Starting from the starting point of the line segment, the window is slid in sequence, with a step size of s each time; for each window, the number of data points in the window that exceed The number of data points c:

[0078]

[0079] Where Count(·) represents statistical calculation; Represents a line segment On the Pressure data points,

[0080] Step 5.2.2: If satisfied Then record the window and the data points contained in the window; if Then calculate the next footage, that is, update the footage index value Switch to the next line segment Execute step 5.2.1 to continue the sliding window traversal and condition judgment;

[0081] Step 5.2.3: Merge all windows that meet the conditions. The endpoints of the windows that meet the conditions are the edges of the predicted significant pressure area. The line segment formed by connecting the endpoints is the predicted significant pressure area line segment. Record the endpoint location.

[0082] Furthermore, the specific process of step 5.4 is as follows:

[0083] Calculate the footage as The area of ​​the significant pressure area predicted at that time The total area S of the predicted significant pressure area is obtained by adding them up. pre,total ;definition For early warning threshold, the area ratio is calculated as When , it is determined that there will be a large area of coming pressure, and an early warning signal is sent in time; if , it is determined that there will be no large area of coming pressure, and no early warning is performed, at this time the footage index value is switched to the next line segment renewed from the starting point of the line segment, step 5.2.1 is executed to perform traversal and condition judgment by using the sliding window method;

[0084] In the prediction of significant coming pressure, a comprehensive prediction index is calculated, and an early warning grading mechanism is designed according to the comprehensive prediction index;

[0085] The formula for calculating the comprehensive prediction index is:

[0086]

[0087] Wherein, F is the comprehensive prediction index; n represents the step distance number of the predicted significant coming pressure area; is a dynamic weight coefficient based on the relationship between the pressure data points and , and the calculation formula is:

[0088]

[0089] The early warning grading mechanism is designed as follows:

[0090] Set the first threshold F1 and the second threshold f2 of the comprehensive prediction index, if f1≤F<F2, it is a first-level early warning, prompting local pressure concentration, and the monitoring frequency needs to be encrypted; if F≥F2, it is a second-level early warning, requiring immediate stop of mining and starting of emergency plan; if F<F1, it is not early warning.

[0091] The beneficial technical effects brought by the present application are as follows.

[0092] High accuracy: by real-time acquisition of hydraulic support pressure data and use of various data processing and analysis methods such as data standardization processing, surface fitting and projection, the characteristics and laws of roof pressure change can be mined. Compared with the traditional method, the significant coming pressure area can be more accurately predicted, the misjudgment and omission are reduced, and more reliable basis is provided for coal mine safety production.

[0093] Dynamic update: during the prediction process, when the footage exceeds a certain distance, the new significant coming pressure area data can be accumulated into the historical data, the relevant characteristic values are recalculated, and the dynamic update of the prediction model is realized. This makes the prediction method better adapt to the changing geological conditions and mining environment in the coal mining process, and always maintains high prediction accuracy.

[0094] Risk early warning: the application of edge detection technology can quickly detect mutations and edge features in data, and identify potential stress concentration areas in advance. Once a large stress area is predicted, an early warning is issued in time to give coal mine workers enough time to take effective preventive measures and reduce the risk of accidents. BRIEF DESCRIPTION OF DRAWINGS

[0095] Figure 1 Flow chart of the stope roof significant pressure prediction method based on edge detection of the present application.

[0096] Figure 2 Roof pressure curve map fitted by data points of the present application.

[0097] Figure 3 Projection area map of the present application.

[0098] Figure 4 Overall division of the coal mining face of the present application.

[0099] Figure 5 Flow chart of the significant pressure area predicted by the edge detection technology of the present application.

[0100] Figure 6 Calculation result diagram of the present application as the working face advances.

[0101] Figure 7 Two-dimensional mine pressure significant area cloud chart of part of the data in time period 1 in experiment 1 of the present application.

[0102] Figure 8 True stress concentration area diagram of part of the data in time period 1 in experiment 1 of the present application.

[0103] Figure 9 Predicted stress concentration area diagram of part of the data in time period 1 obtained by the edge detection method in experiment 1 of the present application.

[0104] Figure 10 Two-dimensional mine pressure significant area cloud chart of part of the data in time period 2 in experiment 2 of the present application.

[0105] Figure 11 True stress concentration area diagram of part of the data in time period 2 in experiment 2 of the present application.

[0106] Figure 12 Predicted stress concentration area diagram of part of the data in time period 2 obtained by the edge detection method in experiment 2 of the present application. DETAILED DESCRIPTION

[0107] The present application will be further described in detail below in combination with the drawings and specific embodiments:

[0108] The present invention introduces edge detection technology into the field of mine roof pressure prediction, and combines it with big data acquisition and processing technology, surface fitting technology, etc. to form a complete prediction system. This multi-technology integration method breaks through the limitations of traditional prediction methods and provides a new idea and method for roof pressure prediction. The present invention also establishes a dynamically updated prediction model, which can continuously incorporate new data into the model for recalculation and analysis as the mining progresses. This dynamic update mechanism makes the prediction model adaptive, able to timely reflect changes in the mine roof pressure, and always maintain the accuracy of the prediction of the pressure area, which is rare in previous roof pressure prediction technologies.

[0109] The present invention is a data-driven method used to predict the distribution and changing trend of future roof pressure based on existing historical data and real-time collected data. Its core lies in mining the laws and features in the data through mathematical and statistical methods, thereby realizing the prediction of unknown situations. Figure 1 As shown, a method for predicting significant roof pressure in a stope based on edge detection includes the following steps:

[0110] Step 1: Obtain the real-time pressure data of the hydraulic support of the electro-hydraulic control system; the specific process is as follows:

[0111] The high-precision pressure sensor equipped with the downhole electro-hydraulic control system is used to collect the pressure data of the hydraulic support in real time according to the set high-frequency sampling frequency f. In the footage interval [x1, x2], the i-th support y perpendicular to the propulsion direction is i The pressure data sequence collected by the upper pressure sensor is recorded as z=[z i1 ,z i2 ,…,z ih ], where h is the number of sampling points in the footage interval, z ih The footage is x h When the i-th bracket y i Finally, the collected pressure data is transmitted in real time through the data transmission network and stored in the database.

[0112] Step 2: Preprocess the pressure data, including outlier detection and processing, missing value processing, and data denoising. The specific process is as follows:

[0113] Step 2.1, outlier detection and processing;

[0114] For the pressure data series, calculate its mean and standard deviation:

[0115]

[0116] Among them, μ iis the mean (average) of the pressure data collected by the pressure sensor on the i-th bracket within a certain period of time; z ij Indicates footage is x j When the i-th bracket y i The pressure value on the track; j is the index value of footage; σ i is the standard deviation of the pressure data collected by the pressure sensor on the i-th bracket within a certain period of time, which is used to measure the fluctuation amplitude or dispersion of these data around the mean; N i is the total number of pressure data points collected by the pressure sensor on the i-th bracket;

[0117] If |z ij -μ i |>3σ i , then determine z ij For outliers, replace them with the median M of the sequence i . M i is the median of all data values ​​of the pressure sensor on the i-th bracket;

[0118] Step 2.2, missing value processing;

[0119] For data points with missing values, linear interpolation is used to fill them. Define the distance corresponding to the missing value as x k , if z ik Missing, and z i(k+1) 、z i(k-1) Known, then:

[0120]

[0121] Among them, z ik 、z i(k+1) 、z i(k-1) Indicates the footage is x k 、x k+1 、x k-1 When the i-th bracket y i The pressure value on the y-axis; k, k+1, k-1 are index values ​​of different footages respectively;

[0122] Step 2.3: Data denoising;

[0123] The cleaned data is denoised using the sliding average filter method. Assuming the sliding window size is m (that is, the number of observation points involved in the calculation within each footage sliding window), each pressure value is denoised using the formula:

[0124]

[0125] Among them, z′ ij Indicates footage is x j When the i-th bracket yi The pressure value after denoising; z il Indicates footage is x l When the i-th bracket y i The pressure value on the y-axis; l is the index value of footage;

[0126] When l<1 or l>N i When the forward filling method is used, that is, z il =z i1 or . z i1 、 Indicates the footage is x1, When the i-th bracket y i The pressure value on.

[0127] Step 3: Use surface fitting technology to fit the data points into a continuous surface and project it to extract the spatial distribution characteristics of the roof pressure. The specific process is as follows:

[0128] Step 3.1, fitting the distribution of pressure data in three-dimensional space by surface fitting technology, and constructing a top plate to press the surface;

[0129] Extract data points from the real-time pressure data of the hydraulic support collected by the electro-hydraulic control system after preprocessing in step 2, and construct a top plate to press the curved surface in three-dimensional space. The top plate to press the curved surface fitted by the data points is as follows: Figure 2 The advancing direction along the working face is set as the x-axis (representing the footage along the advancing direction of the working face, in m), the vertical direction is set as the y-axis (representing the support number perpendicular to the advancing direction, used to identify different hydraulic supports, in units of racks), and the pressure value is set as the z-axis, in units of MPa; Figure 2 The black area in the figure is the top plate pressing the curved surface. j When the i-th bracket y is collected i The pressure value is z ij , then it corresponds to the three-dimensional coordinate (x j ,y i ,z ij ), and thus construct the data point set of the top plate to press the surface:

[0130] {(x j ,y i ,z ij )|i=1,2,...,α;j=1,2,...,β};

[0131] Among them, α is the total number of brackets; β is the total number of footage indexes;

[0132] The top plate presses the surface Z = f(x, y), and its fitting model is:

[0133] Z = a0 + a1x + a2y + a3x + a4y + a5xy 2 2

[0134] Wherein, f(x, y) is the pressure value of the support y collected by the pressure sensor when the footage is x, the unit is megapascal (MPa); a0, a1, a2, a3, a4, a5 are different to be determined coefficients.

[0135] The error function E of the roof pressure curve is constructed for the roof:

[0136]

[0137] The partial derivative of the error function E of the roof pressure curve with respect to the coefficients a0, a1, a2, a3, a4, a5 is taken, and the partial derivative is set to 0, to obtain a system of equations containing 6 equations. Solve the equation set to determine the values of the coefficients a0, a1, a2, a3, a4, a5, and further determine the specific expression of the roof pressure curve Z = f(x, y).

[0138] Step 3.2, projection processing;

[0139] The pressure data value for cutting the curve is set as D0; the plane Cut the determined roof pressure curve, project the curve obtained by cutting to the cutting plane parallel to the xoy plane to obtain the projection area. At the same time, record the data point information in the projection area, including the pressure value, the corresponding support position and the collection time, etc.

[0140] Step 4, calculate the parameters related to the projection area; the specific process is:

[0141] Step 4.1, calculate the area of the projection area and the average value of the data in the area;

[0142] Figure 3 The projection area graph, the numbers 1-20 in the graph are the plane Cut the projection of the cutting plane obtained by cutting the curve into different projection areas, divide each projection area into MxN grids, and each grid has a size of Δx x Δy. Count the number of grids falling within the projection area K, then the area S of the rth projection area is: r

[0143] S r = K x (Δx x Δy);

[0144] Wherein, Δx, Δy are the length and width of each grid respectively; M, N are the number of grids along the x axis and the y axis respectively;

[0145] The total area S of the projection area is the sum of the areas of all the projection areas:

[0146] ​​​

[0147] where q is the number of projection regions, r = 1, 2, …, q;

[0148] Average area of all projection regions is the average of all projection region areas (calculated when multiple projection regions exist, if there is only one projection region then ):

[0149]

[0150] The set of pressure data points of the rth projection region is:

[0151]

[0152] where P rg is the value of the gth pressure data point in the rth projection region; g is the number of pressure data points of the rth projection region.

[0153] For the rth projection region, the average pressure data is:

[0154]

[0155] where P ro is the value of the oth pressure data point in the rth projection region;

[0156] Average data of all projection regions is the average of all pressure data points in all projection regions:

[0157]

[0158] Step 4.2, comparison and adjustment;

[0159] Figure 4 is a schematic diagram of the overall division of the coal mining face, and mining is carried out along the advancing direction of the face, wherein the A region is a historical data region, i.e. the mined region of the face, and the B region is a region to be predicted, i.e. the region to be mined of the face, and subsequent prediction is analyzed in the B region. x1, x2, x3 are different footage.

[0160] Calculate the total area S total of the mined face region, the formula is:

[0161] S total = l push x l ten ;

[0162] wherein, lpush is the historical advancing distance of the working face; l ten is the tendency length of the working face;

[0163] S and S total are compared to determine whether the condition 0.15S total < S < 0.2S total is met. If the condition is not met, the value of D0 needs to be adjusted, and the surface cutting and projection operations are performed again until the condition is met. Then, the data analysis of the historical advancing distance ends, and the average value of the data and the average value of the area of all the projected regions are recorded.

[0164] Step 5, significant pressure area prediction and parameter update. This step uses edge detection technology to quickly detect the mutation characteristics in the pressure data and identify potential pressure concentration areas (significant areas). As the mining progresses, new data is continuously collected and analyzed. When the footage exceeds a certain distance, the new significant pressure area data is accumulated into the historical data, and the relevant characteristic values are recalculated to realize dynamic updating of the prediction model.

[0165] As Figure 5 is the flowchart of using edge detection technology to predict significant pressure areas, the average value of the data and the average value of the area of the significant areas are calculated through historical data analysis, and the possible areas of large pressure are predicted based on this. When the footage reaches a certain distance, the characteristic values of the significant pressure areas can be updated. The specific process is as follows:

[0166] Step 5.1, according to the part of the step distance that has appeared, whether there will be a large pressure appears;

[0167] Assume that the footage of the current prediction stage is is the footage index value of the prediction stage. In order to better show the advancing state of the working face, the mine pressure data points of the current prediction stage are connected into a line segment contains multiple pressure data points is the line segment The first pressure data point on the line segment is the total number of pressure data points on the line segment

[0168] Step 5.2, use the sliding window method to traverse the line segment to determine the predicted significant pressure area line segment and the endpoint position. The specific process is as follows:

[0169] Step 5.2.1, set a fixed length sliding window, the window length is w, and the step length of the sliding window is s ， from ​​Starting from the starting point of the line segment, the window is slid in sequence, with a step size of s each time. For each window, the number of data points in the window that exceed The number of data points c:

[0170]

[0171] Where Count(·) represents statistical calculation; Represents a line segment On the Pressure data points,

[0172] Step 5.2.2: If more than The number of data points c accounts for more than 80% of the total number of data points, that is, Then record the window and the data points contained in the window; if Then calculate the next footage, that is, update the footage index value Switch to the next line segment Execute step 5.2.1 to continue the sliding window traversal and condition judgment;

[0173] Step 5.2.3: Merge all windows that meet the conditions. The endpoints of the windows that meet the conditions are the edges of the predicted significant pressure area. The line segment formed by connecting the endpoints is the predicted significant pressure area line segment. Record the endpoint location.

[0174] Step 5.3: Determine whether the current When satisfied When , the calculation of step 5.4 is performed, that is, only when That is, only after analyzing at least two footage segments can the subsequent predicted area calculation be performed. Then calculate the next footage, that is, update the footage index value Switch to the next line segment Execute step 5.2.1 to continue the sliding window traversal and condition judgment;

[0175] Step 5.4: Calculate the area ratio and set the warning threshold. When the area ratio is greater than the set warning threshold, the warning mechanism is triggered and a warning grading mechanism is designed to perform graded warnings. The specific process is as follows:

[0176] Calculate the footage as The area of ​​the significant pressure area predicted at that time The total area S of the predicted significant pressure area is obtained by adding them up. pre,total ;definition The area ratio is calculated as the pre-warning threshold value When , it is determined that a large coming pressure area will appear, and a pre-warning signal is sent out in time. If , it indicates that the current stage has not yet reached the condition that a large coming pressure area may appear; at this time, the footage index value is switched to the next line segment , and the process is restarted from , the starting point of the line segment, and step 5.2.1 is performed to traverse and judge the conditions using the sliding window method.

[0177] For example, Figure 6 is a schematic diagram of the calculation results as the working face advances. Figure 6 , L salien,1 is the line segment of the predicted significant coming pressure area when the footage is x1; S pre,2 is the area of the predicted significant coming pressure area when the footage is x2; S pre,total is the total area of the predicted significant coming pressure area accumulated when the footage is x3.

[0178] In predicting significant coming pressure, the distribution of mine pressure data points, the relationship between the predicted area and historical data, and other factors are considered to calculate a comprehensive prediction index to measure the degree of prediction of significant coming pressure; the formula for the comprehensive prediction index is:

[0179]

[0180] where F is the comprehensive prediction index; n represents the number of step distances (the distance between adjacent footage) of the predicted significant coming pressure area. is a dynamic weight coefficient based on the relationship between the pressure data points and , and the calculation formula is:

[0181]

[0182] It reflects the proportion of pressure data exceeding the average value at the current step distance, and the higher the proportion, the greater the weight of predicting a significant coming pressure area at this step distance.

[0183] Pre-warning grading mechanism:

[0184] Set the first threshold value F1 and the second threshold value F2 of the comprehensive prediction index. On the basis of the original pre-warning, if F1≤F<F2 is met, it is upgraded to a first-level pre-warning, prompting local pressure concentration, and the monitoring frequency needs to be increased; if F≥F2 is met, it is upgraded to a second-level pre-warning, requiring immediate stop of mining and starting of an emergency plan; if F<F1 is met, there is no pre-warning.

[0185] Step 5.5, if the footage If the distance exceeds 30m, the area and data of the significant pressure area of ​​the predicted part are added to the historical data, and the area mean and data mean of the significant pressure area are recalculated and updated, that is, an update is made. and If the advance If it does not exceed 30m, it means that the data of the current footage stage does not meet the conditions for updating historical data. At this time, update the footage index value Switch to the next line segment After the switch is completed, for the new line segment Repeat the sliding window analysis in step 5.2.

[0186] To verify the effectiveness and reliability of this invention, experiments were conducted using rock pressure data from different time periods at the XV1311 working face. The experimental results show that the edge detection prediction method of this invention can successfully predict pressure concentration areas and change trends that are consistent with actual conditions. The specific experimental data and analysis are as follows:

[0187] Experiment 1: Verification of mine pressure data for time period 1;

[0188] The mine pressure data of the XV1311 working face from cut 1558 (1247.0m) to cut 1605 (1284.6m) from [2025-2-3 9:23:08] to [2025-2-8 1:03:39] were used to identify multiple significant mine pressure areas through surface fitting and projection analysis. Figure 7 Only the two-dimensional significant rock pressure areas of part of the data in time period 1 are shown. At the same time, the area and data mean of each area are calculated, as shown in Table 1:

[0189] Table 1 Data results of significant mine pressure areas for some data in time period 1

[0190]

[0191]

[0192] The edge detection method of the present invention is used to predict the pressure concentration area based on the partial footage of time period 1. Figure 8 、 Figure 9 The actual pressure concentration area and the predicted pressure concentration area of ​​some data in time period 1 are shown respectively. Figure 8 and Figure 9 It can be seen that the method of the present invention predicts the pressure concentration area that is consistent with the actual situation.

[0193] Experiment 2: Verification of mine pressure data for time period 2;

[0194] The mine pressure data of the time period 2 of the working face of XV1311 from the 1604th cutter (1283.8m) to the 1678th cutter (1343.0m) and the time from 【2025-2-8 0:1:54】-【2025-2-14 23:12:14】 are analyzed by surface fitting and projection to identify a plurality of mine pressure significant regions. Figure 10 The two-dimensional mine pressure significant regions of part of the data in the time period 2 are shown. Meanwhile, the area and data average of each region are calculated, as shown in Table 2:

[0195] Table 2: Mine pressure significant region data results of part of the data in the time period 2

[0196]

[0197] The edge detection method of the application is used to predict the pressure concentration region according to part of the footage of the time period 2. Figure 11 、 Figure 12 The real pressure concentration region and the predicted pressure concentration region of part of the data in the time period 2 are respectively shown. From Figure 11 and Figure 12 it can be seen that the pressure concentration region predicted by the method of the application is consistent with the actual situation.

[0198] Of course, the above description is not a limitation of the application, and the application is not limited to the above examples. Changes, modifications, additions or replacements made by the person skilled in the art within the essential scope of the application shall also belong to the protection scope of the application.

Claims

1. A method for predicting significant roof pressure in a stope based on edge detection, characterized in that: It includes the following steps: Step 1: Obtain the real-time pressure data of the hydraulic support in the electro-hydraulic control system; Step 2: Preprocess the pressure data, including outlier detection and processing, missing value processing, and data denoising; Step 3: Fit the data points into a continuous surface through surface fitting technology and project it to extract the spatial distribution characteristics of the roof pressure; Step 4: Calculate the relevant parameters of the projection area; Step 5: Use edge detection technology to predict the significant pressure-increasing area and update the parameters.

2. The method for predicting significant roof pressure in a stope based on edge detection according to claim 1 is characterized in that: In step 1, the pressure sensor equipped with the downhole electro-hydraulic control system is used to collect the pressure data of the hydraulic support in real time according to the set high-frequency sampling frequency f; within a certain footage interval, the i-th support y perpendicular to the propulsion direction is i The pressure data sequence collected by the upper pressure sensor is recorded as Where h is the number of sampling points in the footage interval, z ih Indicates footage is x h When the i-th bracket y i Finally, the collected pressure data is transmitted in real time through the data transmission network and stored in the database.

3. The method for predicting significant roof pressure in a stope based on edge detection according to claim 2 is characterized in that: The specific process of Step 2 is as follows: Step 2.1: Outlier detection and processing; For the pressure data sequence, calculate the mean value and standard deviation: Among them, μ i is the mean value of the pressure data collected by the pressure sensor on the i-th bracket within a certain period of time; z ij Indicates footage is x j When the i-th bracket y i The pressure value on the track; j is the index value of footage; σ i is the standard deviation of the pressure data collected by the pressure sensor on the i-th bracket within a certain period of time; N i is the total number of pressure data points collected by the pressure sensor on the i-th bracket; If |z ij -μ i |>3σ i , then determine z ij For outliers, replace them with the median M of the sequence i ;M i is the median of all data values ​​of the pressure sensor on the i-th bracket; Step 2.2: Missing value processing; For data points with missing values, linear interpolation is used to fill them; the corresponding distance of the missing value is defined as x k , if z ik Missing, and z i(k+1) 、z i(k-1) Known, then: Among them, z ik 、z i(k+1) 、z i(k-1) Indicates the footage is x k 、x k+1 、x k-1 When the i-th bracket y i The pressure value on the y-axis; k, k+1, k-1 are index values ​​of different footages respectively; Step 2.3: Data denoising; The sliding average filter method is used to denoise the cleaned data; the sliding window size is m ， Each pressure value is denoised using the following formula: Among them, z′ ij Indicates footage is x j When the i-th bracket y i The pressure value after denoising; z il Indicates footage is x l When the i-th bracket y i The pressure value on the y-axis; l is the index value of footage; When l<1 or l>N i When the forward filling method is used, that is, z il =z i1 or z i1 、 Indicates the footage is x1, When the i-th bracket y i The pressure value on.

4. The method for predicting significant roof pressure in a stope based on edge detection according to claim 3 is characterized in that: The specific process of Step 3 is as follows: Step 3.1: Fit the distribution of the pressure data in the three-dimensional space through surface fitting technology to construct the roof pressure-increasing surface; Extract data points from the real-time pressure data of the hydraulic support preprocessed in step 2, construct a top plate in three-dimensional space to press the curved surface, set the advancing direction along the working surface as the x-axis, the vertical direction as the y-axis, and the pressure value as the z-axis; if the advance is x j When the i-th bracket y is collected i The pressure value is z ij , then it corresponds to the three-dimensional coordinate (x j ,y i ,z ij ), and thus construct the data point set of the top plate to press the surface: {(x j ,y i ,z ij )|i=1,2,…,α;j=1,2,…,β}? Among them, α is the total number of supports; β is the total number of footage indexes; The roof pressure-increasing surface Z = f(x, y), and the fitting model is: Z=a0+a1x+a2y+a3x 2 +a4y 2 +a5xy; Among them, f(x, y) is the pressure value collected by the pressure sensor of support y at the footage of x; a0, a1, a2, a, a4, a5 are different undetermined coefficients; Construct error function E for the top plate to press the surface ： Take the partial derivatives of the error function E of the roof pressure-increasing surface with respect to the coefficients a0, a1, a2, a3, a4, a5, and set the partial derivatives to 0 to obtain a system of equations containing 6 equations; solve this system of equations to determine the values of the coefficients a0, a1, a2, a3, a4, a5, and further determine the specific expression of the roof pressure-increasing surface Z = f(x, y); Step 3.2: Projection processing; Set the pressure data value for cutting the surface to D0; use the plane Cut the determined top plate to press the curved surface, project the cut curve onto the cutting plane parallel to the xoy plane to obtain the projection area; at the same time, record the data point information in the projection area, including the pressure value, the corresponding bracket position and the acquisition time.

5. The method for predicting significant roof pressure in a stope based on edge detection according to claim 4 is characterized in that: The specific process of Step 4 is as follows: Step 4.1: Calculate the projection area and the data mean value within the area; Each projection area is divided into M×N grids, the size of each grid is Δx×Δy, and the number of grids K falling within the projection area is counted. ， Then the area S of the rth projection area is r for: S r =K×(Δx×Δy); Among them, Δx and Δy are the lengths and widths of each grid respectively; M and N are the numbers of grids along the x-axis and y-axis respectively; The total area S of the projection area is the sum of all projection areas: Among them, q is the number of projection areas, r = 1, 2,..., q; Calculate the mean area of ​​all projected areas The set of pressure data points in the rth projection area for: Among them, P rg is the value of the g-th pressure data point in the r-th projection area; g is the number of pressure data points in the r-th projection area; For the rth projection area, the mean pressure data for: Among them, P ro is the value of the oth pressure data point in the rth projection area; Calculate the mean of the data in all projection areas Step 4.2: Comparison and adjustment; Calculate the total area S of the mined working face area total , the formula is: S total ＝l push ×l ten ； Among them, l push is the historical advancement distance of the working face; l ten is the inclined length of the working face; Combine S with S total Compare and determine whether 0.15S is met total <S<0.2S total If the condition is not met, the value of D0 needs to be readjusted and the surface cutting and projection operations need to be performed again until the condition is met. The data analysis of the historical advancement distance is then completed and the mean value of the data calculated for all projection areas is recorded. and the mean area 6. The method for predicting significant roof pressure in a stope based on edge detection according to claim 5 is characterized in that: The specific process of Step 5 is as follows: Step 5.1: Assume that the footage of the current forecast stage is Index value of footage in the prediction stage; connect the mine pressure data points in the current prediction stage into line segments The line contains multiple pressure data points For line segments On the pressure data points; is a line segment The total number of pressure data points on ; Step 5.2: Use the sliding window method to traverse the line segment Determine the predicted significant pressure area segment and endpoint locations; Step 5.3: Determine whether the current When satisfied When , perform the calculation in step 5.4; if Then calculate the next footage, that is, update the footage index value Switch to the next line segment Execute step 5.2 to continue sliding window traversal and condition judgment; Step 5.4: Calculate the area ratio and set the warning threshold. When the area ratio is greater than the set warning threshold, trigger the warning mechanism and design a warning classification mechanism for classification warning; Step 5.5: If the footage If the distance exceeds 30m, the area and data of the significant pressure area of ​​the predicted part are added to the historical data, and the area mean and data mean of the significant pressure area are recalculated and updated, that is, an update is made. and If the advance If the footage index value does not exceed 30m, update the footage index value. Switch to the next line segment After the switch is completed, for the new line segment Repeat the sliding window analysis in step 5.

2.

7. The method for predicting significant roof pressure in a stope based on edge detection according to claim 6, characterized in that: The specific process of Step 5.2 is as follows: Step 5.2.

1. Set a fixed-length sliding window with a window length of w. ， The step size of the sliding window is s ， from Starting from the starting point of the line segment, the window is slid in sequence, with a step size of s each time; for each window, the number of data points in the window that exceed The number of data points c: Where Count(·) represents statistical calculation; Represents a line segment On the Pressure data points, Step 5.2.2: If satisfied Then record the window and the data points contained in the window; if Then calculate the next footage, that is, update the footage index value Switch to the next line segment Execute step 5.2.1 to continue the sliding window traversal and condition judgment; Step 5.2.3: Merge all windows that meet the conditions. The endpoints of the windows that meet the conditions are the edges of the predicted significant pressure area. The line segment formed by connecting the endpoints is the predicted significant pressure area line segment. Record the endpoint location.

8. The method for predicting significant roof pressure in a stope based on edge detection according to claim 7, characterized in that: The specific process of Step 5.4 is as follows: Calculate the footage as The area of ​​the significant pressure area predicted at that time The total area of ​​the predicted significant pressure area is obtained by adding them up definition is the warning threshold, and the calculated area ratio is when If When it is determined that there is no large pressure area at present, no warning is issued, and the footage index value will be updated at this time Switch to the next line segment Restart Starting from the starting point of the line segment, execute step 5.2.1 and use the sliding window method to perform traversal and condition judgment; When predicting the significant pressure-increasing, calculate the comprehensive prediction index and design a warning classification mechanism according to the comprehensive prediction index; The calculation formula of the comprehensive prediction index is: Among them, F is the comprehensive prediction index; n represents the number of steps in the predicted significant pressure area; Based on pressure data points and The dynamic weight coefficient of the relationship is calculated as follows: Design the warning classification mechanism as follows: Set the first threshold F1 and the second threshold F2 of the comprehensive prediction index. If F1 ≤ F < F2 is satisfied, it is a first-level warning, indicating local pressure concentration and requiring an increased monitoring frequency; if F ≥ F2 is satisfied, it is a second-level warning, requiring immediate stop of mining and activation of the emergency plan; if F < F1 is satisfied, there is no warning.

Images