Study on Critical Dynamic Stress of Coarse-grained Soil Based on Large-scale Dynamic Triaxial Test and Neural Network

A series of large-scale dynamic triaxial tests under different confining pressures, dynamic stresses, and water contents were performed to study the accumulative plastic deformation characteristics, dynamic stability, and critical dynamic stress of a typical coarse-grained subgrade bed soil. The test results demonstrate that the relationship curves between the accumulative plastic strain e of the coarse-grained soil and cycle number N of dynamic loads can be divided into failure, stable, and critical types. A grey relation model was then developed to analyze the correlation characteristics of the e -N curve patterns with respect to the confining pressure, dynamic stress, and water content, where the analytic results show that the curve patterns are significantly affected by the three factors, and they have similar importance in determining the dynamic stability state of the soil. A back-propagation (BP)-neural network model was then established to predict the critical dynamic stress of the coarse-grained soil. The results illustrate that the critical dynamic stress of the coarse-grained soil approximately linearly increases with the increase of the confining pressure but nonlinearly decreases as the water content increases. Finally, an empirical formula considering the effect of confining pressure and water content was proposed to evaluate the critical dynamic stress of the coarse-grained soil based on the test and BP-neural network analytic results. © 2023 Science Press. All rights reserved.