Research

This page highlights some research and projects that I have been involved in the past. Some of them are successfully published as journal articles and conference proceedings, which can be found in the Publications page.

Material Point Method (MPM)

• A High-Performance Multi-Phase Material Point Method for Free-Surface Flow and Saturated Soil
  Collaborators: Prof. Kenichi Soga, Dr. Shyamini Kularathna (UC Berkeley), Prof. Krishna Kumar (UT Austin)
  
  A semi-implicit coupled MPM algorithm is developed to circumvent the numerical instabilities exhibited in the limits of pore-fluid incompressibility and low soil permeability conditions. The fractional-step method is adopted for saturated soil to avoid performance issues while using equal-order interpolation functions for displacement and pressure fields. By implicitly treating the diffusion term, it is shown that a time step size that is independent of the soil permeability can be chosen. The code proposed in the current work is available under an MIT license in Github, within the CB-Geo MPM code. Further enhancement to accurately track the evolution of free-surface and to reduce the computational cost of the linear solver utilizing distributed and shared-memory parallelization are two ongoing works.
  GitHub

• Resolving shear band initiation and propagation with eXtended MPM (XMPM)
  Collaborators: Dr. Yong Liang, Prof. Kenichi Soga
  
  An enhanced XMPM formulation is proposed to simulate the evolution of shear bands and post-failure behaviors with large deformations. A localization search algorithm based on the theory of bifurcation is integrated into the XMPM to predict the initiation and propagation of discontinuity. In addition, a formulation of self-contact is proposed to deal with the dynamic frictional contact mechanism between the generated shear planes. In order to ensure the smoothness of the discontinuity surface during localization propagation, a hybrid implicit–explicit description of discontinuity is considered by employing the level-set method and a point cloud approach.
  GitHub

• Staggered coupling strategies of Implicit MPM and FEM
  Collaborators: Dr. Roland Wüchner, Veronika Singer (TU Munich), Prof. Antonia Larese (University of Padova)
  
  Although MPM has been proven to work robustly for problems involving large deformation materials, the accuracy of particle quadrature is significantly lower than the Gaussian quadrature which is used in the traditional FEM. It, therefore, produces less accurate and efficient solutions when being used to simulate problems with small deformation in comparison to the ordinary FEM. It is always desirable to combine the FEM with MPM to take respective advantages of these two methods, particularly for fluid-structure or soil-structure interaction problems. A staggered coupling framework of implicit MPM and FEM was developed assuming staggered systems with nonconforming discretization. By using a penalty-based method, inhomogeneous Dirichlet boundary conditions can be enforced and the contact forces at the MPM-FEM interface can be approximated. Important coupling components, such as the iterative schemes, convergence accelerators, and mapping tools, are also adopted to enhance the robustness of the coupling.
  GitHub

• Treatment of nonconforming Dirichlet boundary conditions in MPM
  Collaborators: Dr. Roland Wüchner, Veronika Singer (TU Munich), Prof. Antonia Larese (University of Padova)
  
  In many geotechnical engineering applications, material boundaries are often subjected to large displacements and deformation. Under these circumstances, the application of boundary conditions using the MPM becomes a challenging task since material boundaries do not always coincide with the background mesh. During my master’s thesis at TUM and past internships, a formulation of penalty augmentation to impose nonhomogeneous, nonconforming Dirichlet boundary conditions in implicit MPM was proposed. The penalty augmentation is implemented utilizing boundary particles, which can move either according to or independently from the material deformation. Furthermore, releasing-contact boundary conditions, as well as the capability to accommodate slip boundaries, are introduced. The accuracy of the proposed method is assessed in both 2 and 3D cases, by convergence analysis reaching the analytical solution and by comparing the results of nonconforming and classical grid-conforming simulations.
  GitHub

A coupled stabilized ISPH-DEM scheme for fluid-rigid-body simulations

Collaborators: Prof. Mitsuteru Asai, Yi Li (Kyushu University)

With a motivation to simulate water-induced natural hazards accurately, a fluid-structure coupled simulation tool is developed employing particle-based methods. The incompressible Smoothed Particle Hydrodynamics (ISPH) method is chosen to model the fluid flow, whereas the rigid body is modeled by using a particle-based Discrete Element Method (DEM). A stabilized ISPH proposed by Asai et al. 2012 is used as it has been proven to produce a smooth and accurate pressure distribution of free-surface fluid flow with breaking and fragmentation. Upon solving the fluid pressure field through an operator splitting approach, the hydrodynamic forces can be applied onto solid objects through SPH kernel averaging technique. At the same time, the rigid bodies may simultaneously experience contact with other objects and boundaries. Modeled by using DEM, the contact force between these rigid bodies can be approximated by employing either the penalty or the impulse-based methods. Our work has focused on developing an ISPH-DEM coupled algorithm, which exhibits higher accuracy, stability, and efficiency. The developed software has been validated and used for simulations of bridge-washout phenomena, which frequently occur during tsunami, as well as debris flow problems.

Accurate modeling of collision response in DEM and multiple-rigid-body simulations

Collaborators: Prof. Mitsuteru Asai, Yi Li (Kyushu University)

Simulations of multi-body dynamics for computer graphics, 3D game engines, or engineering simulations often involve contact and articulated connections to produce plausible results. As higher accuracy and robustness are continually being sought for engineering purposes, an improved multi-body dynamics simulator based on an impulse method is proposed, specifically an energy-tracking impulse (ETI) algorithm that has been modified to handle particle-discretized rigid-body simulations. To decrease the computational costs of the simulations, a fixed moderate time increment is assumed, allowing multiple-point contacts within a single time increment. The treatment between point-to-point and multiple-point contacts, which include edge-to-surface and surface-to-surface contacts, are, therefore, distinguished through an additional sub-cycling iterations to ensure the conservation of energy.