Implementation Of Multiscale Discrete Element Method Based On Artificial Neural Network

Asphalt pavement is commonly used for its convenience,texture,and ease of construction.Study on asphalt mixtures has also been constrained by analytical tools whereas only empirical methodologies have long been widely used.Comparing with the empirical method,the numerical method has several benefits such as time-saving,low cost,and flexible variable evaluations.Although the finite element method?FEM?is commonly used to work for these challenges on the engineering level,it is hard to accurately simulate the discrete existence between particles and hard to deals with complex material with large numbers of parameters.In contrast,discontinuity-based numerical approaches such as Discrete Element Method?DEM?prove useful in the analysis of micro-structural problems,such as the soil,sand,and asphalt concrete.The multiscale simulation is a tremendous improvement through coupling the DEM and FEM separately at micro and macro-level.This work proposes a methodology for multi-scale simulation by using Finite Element Method?FEM?and Discrete Element Method?DEM?based on Artificial Neural Network?ANN?,to examine and predict the relationship between force-displacement and contact between particle to particle and particle to the wall.The newly concepts are implemented into the basic technique of DEM and FEM in order to create a new model appropriate for analyzing the deformation of the structure during applying displacement or load.A software generated,discrete model of elements was used to simulate asphalt concrete geometry by PFC^2D software,and a finite element analysis was used to simulate the deforming shape of the structure by using ANSYS.Moreover after simulating the model in FEM and DEM,with a set of input parameters displacements?i.e.Ux₁ Uy₁…Ux₄ Uy₄?and output parameters forces?i.e.Fx₁ Fy₁…Fx₄ Fy₄?in x and y directions or both x-y directions were trained an artificial neural network ANN to predict correlation of force-displacement and ensembles the behavior of the structure in MATLAB software.The overall result demonstrated that simulation of DEM and FEM based on ANN could become a useful method to provide a more appropriate and cost-effective framework for force-displacement in the structure.