RBF neural network based creep model of red clay-anchor interface

The creep characteristics of red clay-anchor interface are recognized to dominate essentially the long-term stability of red soil anchored slope. A customized test system for soil-anchor interface shear creep performance is used to study the creep behavior of red clay-anchor interface. The full-process creep curve is obtained by stepwise loading. By using Chen’s methods, the stepwise loading-based full-process creep curves are transformed equivalently to a cluster of creep curves under different stress levels. By selecting the creep results under partial stress levels as learning samples and through supervised learning for those samples, a RBF neural network creep model for red clay-anchor interface is established. Furthermore, another part of the creep results are predicted by the purposed RBF neural network creep model. The results indicates that this model works well in fitting and prediction.