Analytic laws for direct calibration of discrete element modeling of brittle elastic media using cohesive beam model

Abstract

In this study, a new methodology for the calibration of microscopic parameters for the Cohesive Beam Model (CBM) of discrete element method (DEM) applied to elastic brittle material is presented. This method enables the entry of material mechanical values directly into DEM simulations without any calibration steps. Several DEM simulations of tensile tests with different microscopic parameter values were carried out to generate a database of macroscopic parameter responses. This database was analyzed in order to deduce analytic laws by using non linear least square method. To validate the proposed calibration method, DEM simulations, that use the results of the calibrated microscopic parameters, were carried out. The macroscopic responses were compared to theoretical or experimental values. These validation tests were performed separately for two typical brittle elastic materials, i.e., soda-lime glass and alumina, with different shapes/sizes of discrete domain and various boundary conditions. Results between numerical and experimental values are in good accordance regarding the variability induced by this stochastic approach.

Calibration of parallel bond parameters in bonded particle models via physics-informed adaptive moment optimisation

“…The same initial estimation is used in [13,45]. The research in [14] also shows that a linear relationship between the macroscopic tensile strength and the microscopic tensile strength under the conditions that all the packing parameters and deformation parameters are kept the same. Furthermore, larger microscopic bond parameters intuitively will give rise to larger macroscopic strengths.…”

“…Furthermore, larger microscopic bond parameters intuitively will give rise to larger macroscopic strengths. Inspired by this idea and the parametric analysis conducted in [14], the second estimation of microscopic parameters are developed as follows:…”

“…This model can be viewed as a "model of model" and utilises the data obtained by running the simulation model with different sets of parameters to approximate the relationship between inputs (particle-scale parameters) and outputs (macroscopic responses). Many different surrogate models, such as direct fitting method [13][14][15], polynomial response surface [16], artificial neural networks [17,18], Kriging [19], random forest [20], Gaussian process regression and multi-adaptive regression-splines (MARS) [21], are utilised to calibrate DEM parameters. One challenge of these surrogate models is that the accuracy of the surrogate depends on the number and location of sampling points in the available parameter space, whereas the overall computational costs increase when a large number of computational models are required.…”

A hybrid calibration approach to Hertz‐type contact parameters for discrete element models

“…However, the current “resolution” is able to provide useful explanations why a common optimisation algorithm, such as the gradient descent, has difficulty minimising the loss function in order to obtain satisfactory calibration parameters. In addition, the surrogate models are widely used to calibrate DEM parameters . Figures may support an implication that only the surrogate model with a narrow parameter space is capable of predicting macroscopic input parameters with high accuracy, for the Hertz‐type–based DEM models.…”

A Geometric Algorithm to Generate Random Polydisperse Dense Arrangements of Non Over-Lapping Disk Particles

“…The DEM has a wide range of applications. Among the recent, we identify a few: modeling of grinding wheels [Osa et al 2018]; modeling of brittle elastic materials [Nguyen et al 2019]; simulation of ice floes and floating structures [Liu et al 2018]; synthetic rock mass modeling [Vallejos et al 2016]; manufacturing process simulation in the pharmaceutical industry [Yeom et al 2019]; modeling fracture in rock [Azevedo et al 2013]; bi-dimensional simulation of scree-slope dynamics [Bithell et al 2014]; simulation of earthquake surface fault rupture [Taniyama 2017, Garcia 2018; simulation of sheared granular fault systems with a two-dimensional DEM combined with FDEM [Gao et al 2018]; tissue modeling with deformable cells [Gardiner et al 2015]; modeling polyethylene pipes subjected to axial ground movement [Meidani et al 2018]; simulation of damage evolution in coatings under uniaxial substrate tension [Ghasemi et al 2020]; study of the mechanics of two-dimensional nanoparticle assemblies [Marchi et al 2019].…”