Research highlights : : Fabric evolution & shear localization in sand

Landslide onto highway in Taiwan
Fig.1 Massive landslide obstructing highway in Taiwan.

The Role of Fabric Evolution in Strain Localization

Strain localization is an important precursor for major destructive landslides and debris flow (see Fig. 1 on a landslide occurring in Taiwan running onto a major highway in 2010).Fabric anisotropy is amongst the key triggers for strain localization in soils. Existing studies however commonly assume the soil fabric stays constant throughout the formation of localization, while the evolution of fabric and its interaction with the development of shear band are apparently more physically sounded and potentially more important.

Biaxial shear model setup
Fig.2 Modeling of biaxial shear on anisotropic sand.

We recently employed both continuum-based approach and computational multiscale modeling method to investigate the issue, placing an emphatical focus on the role played by fabric and fabric evolution in the initiation and the development of shear band localization in sand. Towards this goad, we simulated a plane-strain biaxial shear test on sand samples with initial cross anisotropy under both smooth or rough boundary conditions (Fig. 2).


FEM Simulation by Continuum Plasticity Approach Accounting for Fabric Evolution

We first attacked the problem by employing FEM and a critical state model accounting for fabric evolution developed recently by us (see Gao et al., 2014) within the framework of continuum plasticity. The typical shear banding initiation, development and final formation are showcased in Fig. 3. Fig. 3b & d show the developing of shear band under smooth (a) and rough (b) boundary with evolving fabric, respectively. Fig. 3e shows a comparison case of rough boundary condition where the soil fabric is fixed unchanged during the entire loading process. Detail of the continuum study can be found in Gao & Zhao, 2013.

shear band localization
Fig.3 Shear banding in sand under different boundary conditions and fabric constraints. a & b, smooth boundary case with evolving fabric; c & d: rough boundary with evolving fabric; e: rough boundary condition with fixed fabric. (see Gao & Zhao, 2013)

Major findings from this continuum FEM study are: (1) The initiation of strain localization can be affected by both structural constraint imposed by the boundary conditions and the presence of soil fabric, whilst the development of shear band is governed by two competing mechanisms: the evolution of fabric and the structural constraint. (2) The structural constraint tends to exert more biased stress on the sand sand, leading to intensified strain localization on the existing shear band(s), while the evolution of fabric may render the material response to become more coaxial with the applied load, helping the sample to resist the external load more optimally to relieve strain localization. (3) Smooth boundary condition leads to a single shear band (Fig. 3b) which may align more closely with the bedding plane (Type b) or orient roughly perpendicular to it (type a). (4) Rough boundary condition leads to cross-shaped shear bands which become more symmetric at large deformation when the fabric is allowed to evolve freely (Fig. 3d). If the fabric is fixed, the cross-shaped shear bands are asymmetric with the first appearing Type a band being dominant (Fig. 3e).

Hierarchical Mulitiscale Simulation Based on Coupled FEM/DEM

RVE packings with different bedding plane for multilscale modeling
Fig.4 Clumped particles (top) used to generate
RVE packings of different bedding plane
(bottom) for multiscale modeling.

We have recently re-examined the same problem in Fig. 2 by our hierarchical multiscale modeling tool based on coupled FEM and DEM (see Research Highlights). To simulate the pre-existing fabric anisotropy, we used clumped particles (shown in Fig. 4: top) to generated RVE packings of different bedding plane (Fig. 4: bottom). The RVEs are then assigned to each Gauss point of the FEM mesh in Fig. 2 uniformly. Since the DEM particles may rearrange themselves during the loading, the multiscale modeling approach naturally capture the evolution of fabric in the simulation.

Fig. 5 presents the developed shear band patterns observed in the smooth boundary case. Being consistent with the continuum modeling results mentioned above, either Type-a (Fig. 5a) or Type-b (Fig. 5b) single shear-band pattern may occur in the multiscale modeling. The distributions of shear strain, void ratio, cumulative particle rotation and fabric anisotropy measured by particle orientation are highly consistent with one another, any of which can be used to identify localization (Fig. 5a-d). However, neither shear stress nor fabric anisotropy defined by contact normals can be used to shown shear localization (see Fig. 5e as compared to others).

Fig. 6 presents the shear band patterns observed in the rough boundary case. Again the multiscale modeling simulations are consistent with the continuum modeling results , with cross-shaped shear-band pattern observed in our modeling. The firstly appeared Type-a band is dominant over the secondary Type-b band in the forming process of the shear bands. Interestingly, the final cumulative rotation at the center of the sample is zero (Fig. 6c & d), due to canceling of initial anti-clockwise rotation induced by the Type-a band and later clockwise rotation caused by the Type-b band.

The multiscale modeling approach also enable us conveniently correlate the macro observation of shear band with their microscopic material response directly. For detail, please refer to the reference provided below on this topic (Zhao and Guo, Géotechnique 2015; Guo and Zhao, 2015).

References:
Shear band patterns obtained by HMM simulation of smooth boundary cases
Fig.5 Multiscale modeling results of shear banding in biaxially sheared sand sample under smooth boundary conditions.
Shear band patterns obtained by HMM simulation of rough boundary cases
Fig.6 Multiscale modeling results of shear banding in biaxially sheared sand sample under rough boundary conditions.