Study of the neural network constitutive models for turfy soil with different decomposition degree

Abstract

The turfy soil is of a special humus soil. The decomposition degree is the main factor on the physical and mechanical properties of turfy soil. To build the turfy soil constitutive model, there are a few shortages such as the calculation cumbersome and low accuracy for parameter value with the method of traditional models. Furthermore, those methods did not reflect the influence of strength that effected by decomposition degree of the turfy soil. In this paper, the relationship of stress-strain with different decomposition degrees of turfy soil was carried out through indoor tests. Based on above experimental results, an improved method, which divided into different zones according to different decomposition degrees of turfy soil and calculated combining with neural network constitutive model is put forward. The result shows that, the neural network of turfy soil has good fitting precision and good generalization ability. It can fully describe the influence of the turfy soil..