Recognition of the Stress-Strain Curve Based on the Local Deformation Measurement of Soil Specimens in the Triaxial Test

A digital image technique was developed and used to measure the deformation distribution over the entire surface of soil specimens in triaxial tests. The measured deformation process shows that the specimen apparently exhibits three states during the test, i.e., pre-failure, in-failure, and post-failure, in correspondence to the different features of its deformation. The deformation feature in each state is then analyzed by the displacement and strain contours. Additionally, the stress level, S, is calculated to determine whether the soil is in failure at a point (representative element volume (REV)) on the surface of the specimen. Next, the failure zone, namely the shear band, was considered to be enveloped by the curve defined by S = 1 on the stress-level contour map. The stress level is calculated based on the strain, Young's modulus and Poisson's ratio. According to analysis of the deformation feature and the failure process, we recognized the following properties: (i) the deformation of the specimen was approximately uniform in the pre-failure state; (ii) failure occurred from a point and developed gradually until the shear band cleaves the specimen; (iii) in the post-failure state, deformation was exclusively due to the blocks of the specimen sliding along the shear band; (iv) the deformation feature in the shear band was quite different from that outside the shear band. In conclusion, the stress-strain curve of the specimen revealed a structural response, not an elementary response, especially in the in-failure and post-failure states, in which the deformation features of different corner points were different, and the observed deformation for the entire specimen may be the combination of local deformations. Therefore, it is not appropriate to build the constitutive model for soil according to the stress-strain curves of the entire deformation process and to take the specimen as a uniform element in the entire process.