The hazards induced by stratified rock mass creep are still one of the major problems that threaten the safety of underground engineering. This paper takes safe construction of underground roadway in Urumqi mining area as the research background. In this study, we mainly adopted rock mechanics experiments to accomplish the research on creep behavior and crack evolution of stratified structural sandstone. Creep deformation characteristics of stratified structural sandstone under different load were revealed; also, we analyzed the reason why a part of rock samples failed but others were not under the same load. Creep behavior and crack evolution of rock samples without stratified structure have significant randomness. The crack evolution and failure characteristics of stratified structural rock samples were mainly manifested as failure along and cutting through structural plane and their combined forms. Creep strain, creep duration, and creep rate of rock samples with stratified structure had a nonlinear relationship with applied load, such as exponential function or logarithmic function. Understanding the evolutionary relationship between the above parameters and load provides a basis for obtaining the creep behavior of stratified rock mass under different load conditions.
The creep and failure mode of stratified coal is complex and challenging to predict. This is highly attributed to the internal stratified structure generated during coal forming. In this study, the inner stratified structure of coal during creep failure was explored. The evolution relationship between the key creep parameters and the inner stratified structure was evaluated through creep tests. The creep failure characteristics of coal under complex stress were determined. A neural network model was established to predict the creep failure modes, and the output results were subjected to smoothing and tracking to reduce the error. The results showed distinct creep characteristics of coal with various types of stratified structure. Key parameters such as creep variable, steady creep state and creep rate exhibited a nonlinear relationship with stress. The dual control neural network model constructed in the study showed high accuracy and stability. This model can be used as a guidance and reference for the study of creep failure mechanism and elucidation of creep behavior prediction of stratified coal.