45 / 2023-05-15 17:20:26

Bedding effect study on the bursting liability of coal and coal-rock combination with different dip angles of bedding

Rock Burst,Bursting Liability,Bedding Effect,PFC3D,Coal-Rock Combination

Rock burst poses a severe threat to coal mining safety, the identification of the coal bursting liability (CBL) is the basis for predicting and preventing rock burst. As a kind of rock with well-developed bedding structure, researches have shown that the physical and mechanical properties as well as the CBL of coal are influenced by its bedding. Existing methods for identifying bursting liability using intact samples have not considered the bedding effects, and may result in discrepancies between the identified bursting liability and actual level.

This paper investigates the bedding effect on the bursting liability of coal and coal-rock, using machine learning model, numerical simulation experiments and theoretical analysis. The research focuses on bedding with different dip angles (0°, 30°, 45°, 60°, 90°), and the results have important theoretical significance for improving the identification methods of bursting liability. The main contents and conclusions of the paper are as follows:

(1) This paper collected 152 sets of CBL test data. 3 data preprocessing methods (min-max normalization, Z-Score normalization, and centering) were pairwise combined with 3 parameter optimization algorithms (genetic algorithm, particle swarm optimization algorithm, and simulated annealing algorithm). By analyzing and comparing the accuracy and stability of 9 LightGBM discrimination models, the best model for identifying the CBL was selected as the PSO-LightGBM model under Z-Score normalization, with F1-score value of 93.6%.

(2) Using PFC3D, this paper conducted uniaxial compression experiments on 3 types of standard cylindrical coal samples (with strong, weak, and none bursting liability) containing bedding with different dip angles, and comparative analyzed the failure modes, microcrack evolution and stress-strain curve of samples. The results showed that when the dip angle of bedding was 0° and 90°, the samples exhibited brittle failure, with microcracks concentrated at the ends and centers of the samples. When the dip angle was 30°, 45°, and 60°, the samples were dominated by shear failure, and the microcracks developed along the bedding planes. The number of microcracks in samples increased as the CBL level decreased, and showed a trend of first increasing and then decreasing as the dip angle increased. As the dip angle increased, the uniaxial compressive strength R_C and impact energy index K_E of the samples showed a trend of first decreasing and then increasing. When R_C and K_E were used in the PSO-LightGBM discrimination model, the results showed that for coal samples with strong bursting liability, the bedding would not affect the CBL level when the dip angle was 0°, 30°, or 90°. However, when the dip angle was 45° or 60°, the CBL level decreased from "strong" to "weak" due to the bedding influence. For coal samples with weak bursting liability, the CBL level would not be affected by the bedding when the dip angle was 0° or 90°. However, when the dip angle was 30°, 45°, or 60°, the CBL level decreased from "weak" to "none".

(3) This paper conducted uniaxial compression experiments on coal-rock combination samples with different dip angles of bedding (coal samples with strong and weak bursting liability). The results showed that when the dip angle was 0° or 90°, the stress-strain curve of samples exhibited an obvious peak and a steep slope after the peak, the microcracks were not oriented along the bedding planes but generated significant particle damage, and the interconnected particles formed large fragmentation blocks. When the dip angle was 30°, 45°, or 60°, the stress-strain curve was relatively flat, without an obvious peak, and was mostly a "bimodal" curve, with damage occurring mainly at the bedding planes and ends of the coal. As the dip angle increased, the R_C and K_E showed a trend of first decreasing and then increasing. According to the results of the PSO-LightGBM discrimination model, when the dip angle was 0° or 90°, the bursting liability level of the coal-rock samples was not affected by the bedding. However, when the dip angle was 30°, 45°, or 60°, the bursting liability level of the samples was reduced to varying degrees due to the bedding influence.

The above results show that both coal and coal-rock with bursting liability have different degrees of bedding effect. When identifying the bursting liability, the influence of the bedding and its angle should be considered. Research can be done to explore the mechanical parameters of samples and correction methods for the bursting liability index under different bedding angles.