Method for predicting rock burst of top plate of steeply inclined coal seam

Abstract

The invention discloses a steeply inclined coal seam roof rock burst prediction method, which comprises the steps of establishing a 1DCNN-LSTM mine earthquake monitoring model, inputting 10 index data into the model for data processing according to 10 common rock burst risk indexes in the mining industry, and obtaining a data model result of each index; comparing the processed index data with a judgment standard, and judging whether the result of each index is dangerous or not; and judging all results to obtain which indexes are dangerous, giving out early warning, and predicting the rock burst danger level according to the number of early warning indexes. According to the method, the rock burst is predicted through the 1DCNN-LSTM model based on the mine earthquake monitoring model of the one-dimensional convolutional neural network (1DCNN) and the long short-term memory network (LSTM), the prediction accuracy and reliability are improved, rock burst analysis is carried out in combination with real-time monitoring data, potential dangers are found in time, and safety accidents are reduced.

Description

Technical Field

[0001] The invention belongs to the technical field of mining engineering, and in particular relates to a method for predicting rock burst pressure of a steeply inclined coal seam roof. Background Art

[0002] Coal is one of the main energy sources in my country, and its resource conditions are complex. With the increase in the depth of coal mining and the increase in mining intensity, coal mining inevitably faces many scientific and technical problems such as roof disasters, rock burst, coal and gas outbursts, water disasters, and efficient gas extraction. Rock burst prediction technology is one of the more difficult problems to solve.

[0003] Rock burst is a common geological disaster in coal mining, which may lead to serious safety accidents. Traditional rock burst prediction methods usually rely on simple geological analysis and empirical judgment, lack of accurate calculation models, resulting in low prediction accuracy. Summary of the invention

[0004] In view of the above deficiencies in the prior art, the technical problem solved by the present invention is to provide a method for predicting rock burst pressure in the roof of a steeply inclined coal seam. The rock burst pressure is predicted by a 1DCNN-LSTM model based on a mine earthquake monitoring model of a one-dimensional convolutional neural network (1DCNN) and a long short-term memory network (LSTM), thereby improving the accuracy and reliability of the prediction. The rock burst pressure is analyzed in combination with real-time monitoring data to timely discover potential dangers and reduce the occurrence of safety accidents.

[0005] In order to solve the above technical problems, the present invention is implemented by the following technical solutions:

[0006] The present invention provides a method for predicting rock burst pressure on a steeply inclined coal seam roof, which specifically comprises the following steps:

[0007] Step 1: Establish a 1DCNN-LSTM mine earthquake monitoring model. According to the 10 rock burst hazard indicators commonly used in the mining industry, input the 10 indicator data into the model for data processing to obtain the data model results for each indicator.

[0008] Step 2: Compare the processed indicator data with the judgment criteria to determine whether the results of each indicator are dangerous;

[0009] Step 3: After all the results are judged, it is determined which indicators are dangerous, and an early warning is issued. The danger level of rock burst is predicted based on the number of early warning indicators.

[0010] The preferred 10 indicators are: coal seam thickness (X1), coal seam inclination (X2), burial depth (X3), geological structure (X4), coal seam inclination change (X5), coal seam thickness change (X6), gas concentration (X7), roof management (X8), pressure relief method (X9) and coal blasting times (X1). 10 ).

[0011] Furthermore, the main operation process of the mine earthquake monitoring model of 1DCNN-LSTM: the forget gate (f), input gate (i), and output gate (o) units prevent the forgetting and memory of information. Different information gates have different information structures at a certain moment. The amount of information on the memory cell of the forget gate at time t is:

[0012] ft=δ(Wfhht-1+Wfxxi)

[0013] Where, δ is the Sigmoid nonlinear function; Wfh is the weight matrix of the forget gate and the output at the previous moment; ht-1 is the output information at the moment t-1; Wfx is the weight matrix of the forget gate and the input at the previous moment; xi is the input information at the moment x;

[0014] The amount of information on the memory cell at the input gate at time t:

[0015] it=δ(Witht-1+Wixxi)

[0016] Where, Wit is the weight matrix of the input gate and the output at the previous moment; Wix is ​​the weight matrix of the input gate and the input at the previous moment;

[0017] The amount of information on the memory cell at the output gate at time t:

[0018] ot=δ(Wohht-1+Waxxi)

[0019] Where, Woh is the weight matrix between the output gate and the output at the previous moment; Wax is the weight matrix between the output gate and the input at the previous moment;

[0020] The amount of output information at time t:

[0021] ht=ot⊙tanhct

[0022] In the formula, ot—output gate determines the amount of information on the memory cell at time t; ⊙—vector dot multiplication operation; ct—the amount of information on the memory cell at time t.

[0023] From the above, the method for predicting rock burst in steeply inclined coal seam roof of the present invention has at least the following beneficial effects:

[0024] 1. Improve prediction accuracy: The 1DCNN-LSTM model is used to predict rock burst, which improves the accuracy and reliability of the prediction.

[0025] 2. Real-time monitoring and analysis: Combining real-time monitoring data with rock burst analysis can timely identify potential dangers and reduce the occurrence of safety accidents.

[0026] 3. Decision support: It provides scientific basis and decision support for rock burst prevention and control technology in coal mining process.

[0027] 4. The present invention combines the relevant data of 10 rock burst hazard indicators commonly used in the mining industry to establish a double-layer Stacking fusion rock burst hazard level prediction model. At the same time, based on the real-time monitoring of the stress time series data in the area, combined with the instability data index of the coal body in the destruction zone and the deformation energy index released in the elastic zone that causes the coal body to be thrown out in the form of power, a coal tunnel rock burst prediction model is established. On the basis of the LSTM prediction model, the 1DCNN-LSTM model is established using the two indicators of mean absolute error and root mean square error to evaluate the prediction performance of the model. It is used to predict the evolution trend of the surrounding rock stress in the study area. Compared with the traditional mine earthquake detection model, the 1DCNN-LSTM model has a higher prediction accuracy for the evolution trend of the stress. The present invention can provide new methods and approaches for the identification of rock burst hazard levels, monitoring and prediction of microseisms and stresses in the process of coal mining, which is helpful to adopt targeted rock burst prevention and control technologies, and has certain guiding significance for the safe mining of coal mines.

[0028] 5. For the prediction of rock burst in steeply inclined coal seam roof, the model comprehensively utilizes the relevant data of 10 rock burst hazard indicators commonly used in mining, and establishes a double-layer stacking fusion rock burst hazard level prediction model. In addition, based on the real-time monitoring of stress time series data, combined with the instability data indicators of the coal body in the destruction zone and the deformation energy indicators released in the elastic zone, a coal tunnel rock burst prediction model was established.

[0029] 6. The 1DCNN-LSTM model is used to predict the evolution trend of surrounding rock stress in the study area. The prediction performance of the model is evaluated by two indicators: mean absolute error and root mean square error. Compared with the traditional mine earthquake detection model, the 1DCNN-LSTM model shows higher accuracy in predicting the stress evolution trend. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the drawings of the embodiment are briefly introduced below.

[0031] Figure 1 This is a diagram showing the results of the steps of data processing using the model in the present invention;

[0032] Figure 2 Run the flow chart for the system. DETAILED DESCRIPTION

[0033] The following is a detailed description of the specific embodiments of the present invention in conjunction with the accompanying drawings, which is a part of this specification and illustrates the principles of the present invention through examples. Other aspects, features and advantages of the present invention will become clear from the detailed description. In the referenced drawings, the same or similar components in different figures are represented by the same reference numerals.

[0034] The embodiment of the present invention comprises the following steps:

[0035] Data collection: Collect relevant geological data and real-time monitoring data of steeply inclined coal seams.

[0036] Model establishment: The rock burst prediction model is established using the 1DCNN-LSTM structure, combined with the rock burst hazard indicators commonly used in the mining industry.

[0037] Model training and testing: Train the model using historical data and evaluate the model’s predictive performance on an independent test set.

[0038] Real-time prediction and analysis: Use the trained model to analyze real-time data, predict the possibility of rock burst, and provide guidance for rock burst prevention and control.

[0039] like Figure 1 and Figure 2 As shown, the present invention proposes a method for predicting rock burst pressure based on a mine earthquake monitoring model of 1DCNN-LSTM. The rock burst pressure prediction method of the steeply inclined coal seam roof of the present invention comprises the following steps:

[0040] Step 1: Establish a 1DCNN-LSTM mine earthquake monitoring model. According to the 10 commonly used rock burst hazard indicators in the mining industry, input the 10 indicator data into the model for data processing to obtain the data model results for each indicator. The 10 indicators are: coal seam thickness (X1), coal seam inclination (X2), burial depth (X3), geological structure (X4), coal seam inclination change (X5), coal seam thickness change (X6), gas concentration (X7), roof management (X8), pressure relief method (X9) and number of coal blasting (X1). 10 ).

[0041] Step 2: Compare the processed indicator data with the judgment criteria to determine whether the results of each indicator are dangerous.

[0042] Step 3: After all the results are judged, it is determined which indicators are dangerous, and an early warning is issued. The danger level of rock burst is predicted based on the number of early warning indicators.

[0043] The present invention can be used in more complex work sites where traditional prediction methods cannot meet the requirements. The method of using the 1DCNN-LSTM model to predict rock burst pressure on the roof of a steeply inclined coal seam has the following advantages:

[0044] (1) Improve prediction accuracy: This method combines a one-dimensional convolutional neural network (1DCNN) and a long short-term memory network (LSTM) to accurately predict the stress change trend of the surrounding rock in the study area. Compared with the traditional mine earthquake detection model, the 1DCNN-LSTM model shows higher accuracy in predicting the stress evolution trend of rock burst.

[0045] (2) Real-time monitoring and analysis: This prediction method can be combined with real-time monitoring of stress time series data to analyze the development trend of rock burst in real time, discover potential dangers in a timely manner, and thus reduce the occurrence of safety accidents.

[0046] (3) Decision support: Providing scientific basis and decision support will help to take targeted measures to prevent and control rock burst and improve safe mining conditions in coal mines.

[0047] (4) Comprehensive evaluation: This method comprehensively utilizes the impact rock pressure hazard indicators commonly used in the mining industry, real-time monitoring data, coal body instability data indicators, and deformation energy indicators released in the elastic zone to provide comprehensive data support for prediction.

[0048] (5) Technological advancement: The use of advanced deep learning technologies, such as 1DCNN-LSTM combined with double-layer stacking fusion technology, has improved the model's predictive ability and data processing depth.

[0049] The main operation process of the mine earthquake monitoring model of 1DCNN-LSTM: The forget gate (f), input gate (i), and output gate (o) units prevent the forgetting and memory of information. Different information gates have different information structures at a certain moment. The amount of information on the memory cell of the forget gate at time t is:

[0050] ft=δ (Wfhht-1+Wfxxi) (1)

[0051] Where, δ is the Sigmoid nonlinear function; Wfh is the weight matrix of the forget gate and the output at the previous moment; ht-1 is the output information at the t-1 moment; Wfx is the weight matrix of the forget gate and the input at the previous moment; xi is the input information at the x moment.

[0052] The amount of information on the memory cell at the input gate at time t:

[0053] it=δ(Witht-1+Wixxi) (2)

[0054] Where Wit is the weight matrix between the input gate and the output at the previous moment; Wix is ​​the weight matrix between the input gate and the input at the previous moment.

[0055] The amount of information on the memory cell at the output gate at time t:

[0056] ot=δ(Wohht-1+Waxxi) (3)

[0057] Where, Woh is the weight matrix between the output gate and the output at the previous moment; Wax is the weight matrix between the output gate and the input at the previous moment.

[0058] The amount of output information at time t:

[0059] ht=ot⊙tanhct (4)

[0060] In the formula, ot—output gate determines the amount of information on the memory cell at time t; ⊙—vector dot multiplication operation; ct—the amount of information on the memory cell at time t.

[0061] Formula (1) to Formula (4) respectively represent the information composition of the forget gate, input gate and output gate at time t and the amount of information output by the LSTM network model at time t. After a large amount of data training, the LSTM network changes the weights of neurons at each level, further reducing the error between data and achieving the purpose of improving data accuracy. It can be seen that LSTM data is a high-precision network model based on massive data.

[0062] The above is only a preferred embodiment of the present invention, which certainly cannot be used to limit the scope of rights of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and changes can be made without departing from the principle of the present invention, and these improvements and changes are also regarded as the protection scope of the present invention.

Claims

1. A method for predicting rock burst pressure on steeply inclined coal seam roof. It is characterized in that The following steps are involved: Step 1: Establish a 1DCNN-LSTM mine earthquake monitoring model. According to the 10 rock burst hazard indicators commonly used in the mining industry, input the 10 indicator data into the model for data processing to obtain the data model results for each indicator. Step 2: Compare the processed indicator data with the judgment criteria to determine whether the results of each indicator are dangerous; Step 3: After all the results are judged, it is determined which indicators are dangerous, and an early warning is issued. The danger level of rock burst is predicted based on the number of early warning indicators.

2. The method for predicting rock burst in the steeply inclined coal seam roof according to claim 1, It is characterized in that The 10 indicators are: coal seam thickness (X 1 ), coal seam inclination (X 2 )、Burial Depth(X 3 ), geological structure (X 4 ), coal seam inclination change (X 5 ), coal seam thickness change (X 6 ), gas concentration (X 7 ), roof management (X 8 )、Pressure relief method (X 9 ) and coal gun times (X 10 ).

3. The method for predicting rock burst in steeply inclined coal seam roof according to claim 1, It is characterized in that The main operation process of the mine earthquake monitoring model of 1DCNN-LSTM: The forget gate (f), input gate (i), and output gate (o) units prevent the forgetting and memory of information. Different information gates have different information structures at a certain moment. The amount of information on the memory cell of the forget gate at time t is: ft=δ(Wfhht-1+Wfxxi) Where, δ is the Sigmoid nonlinear function; Wfh is the weight matrix of the forget gate and the output at the previous moment; ht-1 is the output information at the moment t-1; Wfx is the weight matrix of the forget gate and the input at the previous moment; xi is the input information at the moment x; The amount of information on the memory cell at the input gate at time t: it=δ(Witht-1+Wixxi) Where, Wit is the weight matrix of the input gate and the output at the previous moment; Wix is ​​the weight matrix of the input gate and the input at the previous moment; The amount of information on the memory cell at the output gate at time t: ot=δ(Wohht-1+Waxxi) Where, Woh is the weight matrix between the output gate and the output at the previous moment; Wax is the weight matrix between the output gate and the input at the previous moment; The amount of output information at time t: ht=ot⊙tanhct In the formula, ot—output gate determines the amount of information on the memory cell at time t; ⊙—vector dot multiplication operation; ct—the amount of information on the memory cell at time t.

Images