Dynamic behavior of subgrade soils greatly affects the serviceability of a railway/highway line, where the critical dynamic stress (σ_cri) mainly determines the dynamic stability of a subgrade. In this study, a series of large-scale dynamic triaxial tests on saturated specimens with different confining pressures (σ₃) and dynamic deviatoric stresses (σ_d) were performed to investigate the dynamic stability of a coarse-grained subgrade soil. A Grey Relational Analysis was first established, where the analyzed results demonstrating that the dynamic stability of the coarse-grained soil (CGS) strongly correlated with σ₃ and σ_d. The accumulative plastic strain (ε_p) versus cycle number (N) curve patterns (i.e., failure, critical, and stable) of the tested specimens were then analyzed using the back-propagation (BP)-neural network, illustrating that the latter can effectively reflect the variation trend of ε_p with respect to σ₃ and σ_d. Finally, an empirical method was proposed to predict the critical dynamic stress of the CGS, and a shakedown limit criterion was introduced to analyze the dynamic strength properties of the CGS under shakedown limit states, revealing that the critical dynamic stress approximately linearly increased with increasing confining pressure.

路基土的动态行为有很大影响铁路/高速公路线，其中临界动应力（的适用性σ _CRI）主要决定路基的动态稳定性。在这项研究中，在饱和的试样的一系列大型动态三轴试验用不同的围压（σ ₃）和动态偏应力（σ _d进行）调查一个粗粒度路基土的动态稳定性。灰色关联分析最初成立，其中，分析结果证明该粗粒土的动态稳定性（CGS）强关联与σ ₃和σ _d。累积塑性应变（ε _p）对循环数（Ñ）曲线图案（即，故障，关键的，并且稳定），那么测试的试样均使用反向传播（BP）-neural网络分析，其示出的是，后者可以有效地反映的变化趋势ε _p相对于σ ₃和σ _d。最后，提出了一种经验方法来预测CGS的临界动应力，并引入了一个减振极限准则来分析CGS在减振极限状态下的动态强度特性，揭示出临界动应力随着围限的增加近似线性增加压力。