针对一维动态节理角度和轴向静载差异的煤岩结构面承载损伤下应力波透射问题，采用室内实验、理论分析和计算机仿真等方法，揭示了界面倾角与轴向静载对煤岩结构面透射应力波的作用机制，结合Barton-Bandis节理本构模型、UDEC离散元仿真与灰狼算法优化BP神经网络技术，对分离式霍普金森压杆(Split Hopkinson Pressure Bar，SHPB)实验过程中应力波透射进行仿真和机器学习，在充分考虑各项参数的基础上进行了显著性正交试验和多因素方差分析，筛选出了主要影响因素并确定了修正方案，得到了煤岩结构面轴向静载与倾角差异下Barton-Bandis本构数值模拟参数修正的机器学习模型，极大地提升了煤岩结构面受冲击状态下变形行为与本构参数之间关联机制的计算效率。研究结果表明：基于BP人工神经网络技术的机器学习预测模型具有良好的适用性，可快速确定当前煤岩结构面倾角与轴向静载下模型参数，提供了一种基于数据驱动的煤岩结构面Barton-Bandis本构模型参数的高效修正方法，同时可以预测给定训练样本以外更大倾角与轴向静载范围下煤岩结构面数值模拟参数。

Given the one-dimensional dynamic joint angle and axial static load difference of the coal rock structural plane under the bearing damage of the stress wave transmittance problem, the mechanism of interface inclination and axial static load on the transmitted stress wave of the coal-rock structural surface was revealed by using indoor experiments, theoretical analysis and computer simulation. The simulation and machine learning of stress wave transmission in the experimental process of Split Hopkinson Pressure Bar (SHPB) were carried out by combining the Barton-Bandis nodal ontology model, UDEC discrete element simulation and Gray Wolf Algorithm optimized BP neural network technology. The significance orthogonal test and multi-factor analysis of variance were carried out with full consideration of the various parameters, which screened out the main influencing factors and determined the correction scheme. Simultaneously, a machine learning model for the correction of Barton-Bandis intrinsic numerical simulation parameters under axial static loading and inclination differences of coal rock structural planes is obtained, which greatly improves the computational efficiency of the correlation mechanism between deformation behavior and intrinsic parameters in the impacted state of coal rock structural planes. This study demonstrates that the machine learning prediction model based on BP artificial neural network technology has well-applicability, which can quickly determine the model parameters under the current inclination angle and axial static load of the coal rock structural plane, provide an efficient data-driven correction method for the parameters of the Barton-Bandis intrinsic model of the coal rock structural plane and also predict the parameters of numerical simulation of the coal rock structural plane under the larger inclination angle and axial static load ranges other than the given training samples.