We perform a detailed analysis of the contact force network in a dense confined packing of pentagonal particles simulated by means of the contact dynamics method. The effect of particle shape is evidenced by comparing the data from pentagon packing and from a packing with identical characteristics except for the circular shape of the particles. A counterintuitive finding of this work is that, under steady shearing, the pentagon packing develops a lower structural anisotropy than the disk packing. We show that this weakness is compensated by a higher force anisotropy, leading to enhanced shear strength of the pentagon packing. We revisit "strong" and "weak" force networks in the pentagon packing, but our simulation data provide also evidence for a large class of "very weak" forces carried mainly by vertex-to-edge contacts. The strong force chains are mostly composed of edge-to-edge contacts with a marked zig-zag aspect and a decreasing exponential probability distribution as in a disk packing.

a b s t r a c tBy means of contact dynamics simulations, we investigate a dense packing composed of irregular polyhedral particles under quasistatic shearing. The effect of particle shape is analyzed by comparing the polyhedra packing with a packing of similar particle size distribution but with spherical particles. The polyhedra packing shows higher shear stress and dilatancy but similar stress-dilatancy relation compared to the sphere packing. A harmonic approximation of granular fabric is presented in terms of branch vectors (connecting particle centers) and contact force components along and perpendicular to the branch vectors. It is found that the origin of enhanced shear strength of the polyhedra packing lies in its higher force anisotropy with respect to the sphere packing which has a higher fabric anisotropy. Various contact types (face-vertex, face-face, etc.) contribute differently to force transmission and fabric anisotropy. In particular, most face-face contacts belong to strong force chains along the major principal stress direction whereas vertex-face contacts are correlated with weak forces and oriented on average along the minor principal stress direction in steady shearing.

Discrete element simulation provides some insight into the alteration of railway ballast after repeated train passings. The present Part 1 is devoted to a 2D model of this granular layer interposed between the deformable ground and the rail sleeper, to which a large number of loading cycles is applied. Ballast grains are modelled as indeformable polygonal solids. A detailed account of the application to this frictional dynamical problem of the Non-Smooth Contact Dynamics numerical method is given. Validation is obtained through comparison with physical experiments performed on assemblies of prismatic mineral grains. Numerical results on the settlement of a track submitted to 20,000 loading cycles or more are presented.

No abstract

For the French conventional railway lines, a layer, namely interlayer, was created in the substructure mainly by the interpenetration of ballast grains and subgrade soils. The in situ investigation indicated that the content of coarse grains decreases over depth. In this study, the resilient modulus and damping ratio of the unsaturated lower part of the interlayer soil were investigated at six different volumetric inclusion contents fv (volumetric ratio of dry coarse grains to the whole sample) by carrying out cyclic triaxial tests following a multi-step loading procedure. The results show that the two dynamic parameters (resilient modulus and damping ratio) change significantly at the beginning of loading and the variation rate decreases as the number of cycles increases. Besides, a mean characteristic volumetric inclusion content fv–cha ranging from 27.8% to 28.9% was identified, separating two zones with different inclusion effects. To verify this observation, X-ray microcomputed tomography (μCT) scans were conducted on as-compacted samples. The results obtained strongly support the existence of fv–cha: when fv ≤ fv–cha, the matrix of fines constitutes the skeleton of the sample with inclusions floating in it, leading to slight changes of the two dynamic parameters with fv. By contrast, when fv > fv–cha, the inclusions dominate the skeleton of the sample, giving rise to significant changes of the two dynamic parameters with fv.