Abstract
A data-driven framework was used to predict the macroscopic mechanical behavior of dense packings of polydisperse granular materials. The discrete element method, DEM, was used to generate 92,378 sphere packings that covered many different kinds of particle size distributions, PSD, lying within 2 particle sizes. These packings were subjected to triaxial compression and the corresponding stress–strain curves were fitted to Duncan–Chang hyperbolic models. An artificial neural network (NN) scheme was able to anticipate the value of the model parameters for all these PSDs, with an accuracy similar to the precision of the experiment and even when the NN was trained with a few hundred DEM simulations. The estimations were indeed more accurate than those given by multiple linear regressions (MLR) between the model parameters and common geotechnical and statistical descriptors derived from the PSD. This was achieved in spite of the presence of noise in the training data. Although the results of this massive simulation are limited to specific systems, ways of packing and testing conditions, the NN revealed the existence of hidden correlations between PSD of the macroscopic mechanical behavior.
Highlights
The discrete element method, DEM, was used to generate 92,378 sphere packings that covered many different kinds of particle size distributions, PSD, lying within 2 particle sizes. These packings were subjected to triaxial compression and the corresponding stress–strain curves were fitted to Duncan–Chang hyperbolic models
The estimations were more accurate than those given by multiple linear regressions (MLR) between the model parameters and common geotechnical and statistical descriptors derived from the PSD
Motivated by the apparent lack of correlation between PSD descriptors and the Duncan–Chang model parameters evidenced in the previous section, in this work we present, as an accurate alternative, the use of neural network (NN) for inferring the macroscopic mechanical behavior of polydisperse granular packings
Summary
The specific values of properties such as strength, compressibility and permeability of dry and cohesionless coarse grain materials (including sand, gravel, railway ballast or rockfill) depend on the features of the constituent particles (intrinsic properties) and on the way in which the particles are arranged (state parameters). The usability of results is limited to specific systems and testing protocols, the approach may be useful for other cases and may shed light on the mechanical behavior of dry coarse-grain soils This is a very timely moment since techniques such as computer vision [51] or X-ray tomography are allowing for an exhaustive characterization of the microstructure of granular packings (including PSD and fabric) [69]. These virtual samples shew the typical behavior of loose sands in triaxial compression but the stress versus strain curves changed from one case to another The results of these experiments are not intended for other systems or tests, but they are used to illustrate a methodology based on NNs. The results of these experiments are not intended for other systems or tests, but they are used to illustrate a methodology based on NNs These are simplified models, far from real soils and not capable for accounting for relevant aspects affecting the mechanical behavior (such as the packing ratio or the particle shape), they still exhibit complexity.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
