| CPC G06F 30/23 (2020.01) [E21C 39/00 (2013.01); G06F 30/27 (2020.01); G06N 3/04 (2013.01); G06N 3/088 (2013.01); G06F 2111/10 (2020.01)] | 4 Claims |
|
1. A method for quickly optimizing key mining parameters of an outburst coal seam, comprising the following steps:
1) constructing an automatic numerical simulating model by constructing a repetitive command system; and obtaining a graphic basic information model of a coal mine, wherein the graphic basic information of the coal mine comprises an inclination angle and a thickness of a coal rock layer; and each coal rock layer is divided into a certain number of units;
2) giving coal mine characteristic information to units in different positions in the graphic basic information model, wherein the coal mine characteristic information comprises initial physical and mechanical parameters; and physical and mechanical property parameters comprise rock bulk density, compression resistance, shear resistance, tensile strength and tangential stiffness;
3) inputting coal mine simulating mining information to perform programmed mining simulation; and outputting and storing stress field changes, displacement field changes, gas pressure field changes and fissure field changes during mining process in a database, wherein the mining information comprises a coal seam mining sequence, mining height and a mining rate;
4) changing the coal seam mining sequence, mining height and mining speed respectively, repeating the step 3), and obtaining a data set of displacement fields, stress fields, gas fields and fracture fields under different coal seam mining sequences, different mining height and different mining speed;
5) changing the initial physical and mechanical parameters of the coal rock layer and repeating steps 2) to 4) to obtain a secondary database of sample data for numerical simulation of the coal layer under different initial physical and mechanical parameters;
6) changing an inclination angle and a thickness of the coal rock layer and repeating steps 1) to 5) to obtain the secondary database of sample data for coal rock simulation of the coal rock layer under different inclination angles and thicknesses;
7) normalizing the sample data for coal rock simulation of the coal rock layer, wherein grey relational analysis is carried out to analyze variable characteristics of the coal seam mining sequence, the mining height, the mining speed, a coal rock layer inclination angle, a coal rock layer thickness and physical and mechanical parameters;
8) constructing and training a CNN-LSTM predicting model, wherein the CNN-LSTM predicting model comprises two convolutional neural layers and a long- and short-term neural layer; and each convolutional neural layer comprises an input layer, a convolutional layer, a pooling layer, a weighted connecting layer, and an output layer, and the CNN-LSTM predicting model comprises a convolutional neural network CNN part and a long and short-term memory network LSTM part; the convolutional neural network CNN part is used to express and extract data at the bottom, and the long and short-term memory network LSTM part is used to receive the output of the CNN extracted features, wherein the CNN-LSTM predicting model predicts stress field changes, displacement field changes, gas field changes and fissure field changes;
9) inputting into the CNN-LSTM predicting model required simulated coal mine information, changing the coal seam mining sequence, mining height and mining speed, and obtaining the stress field changes, the displacement field changes, the gas field changes and the fissure field changes under different mining conditions, wherein the required simulated coal mine information comprises variable characteristics of the inclination angle of the coal rock layer, the thickness of the coal rock layer and the physical and mechanical parameters;
10) taking the changes of the coal seam stress fields, displacement fields, gas fields and fissure fields as four variables to construct a system of ordinary differential equations, wherein a Lorenz differential equation is used to construct a nonlinear prediction model of a gas outburst chaotic system related to the gradient of the gas pressure and the gradient changes of ground stress field; and the differential equation is solved iteratively using a fourth-order Runge-Kutta algorithm;
11) checking the sensitivity of stress field changes, displacement field changes, gas field changes, and fissure field changes to the gradient of gas pressure and the gradient changes of the ground stress fields; and constructing a Lorenz's chaotic primer, constructing an attraction point of coal and gas outburst in coal mining, and using the obtained database to standardize, train and detect the chaotic primer;
12) making the model in the specified stress field change, deformation field change, gas field change and fissure field change, inputting the gradient of the gas pressure and the gradient limit information of the ground stress fields, and repeating the input of the stress field changes, the deformation field changes, the gas field changes and the fissure field changes, and outputting the experimental results, so that every result surrounds Lorenz's chaotic primer, all approaching the actual simulation; and obtaining a chaotic model for predicting coal and gas outburst in a final mature period; and
13) constructing machine autonomous learning, using unsupervised learning methods to continuously learn, and finally getting a smart coal and gas predicting intelligent system.
|