Yaoxin Zhang

2D Nearly orthogonal mesh generation with controls on distortion function

A method to control the distortion function of the Ryskin and Leal (RL) orthogonal mesh generation system is presented. The proposed method considers the effects from not only the local orthogonal condition but also the local smoothness condition (the geometry and the mesh size) on the distortion function. The distortion function is determined by both the scale factors and the averaged scale factors of the constant mesh lines. Two adjustable parameters are used to control the local balance of the orthogonality and the smoothness. The proposed method is successfully applied to several benchmark examples and the natural river channels with complex geometries.

Parallelizing Alternating Direction Implicit Solver on GPUs

We present a parallel Alternating Direction Implicit (ADI) solver on GPUs. Our implementation significantly improves existing implementations in two aspects. First, we address the scalability issue of existing Parallel Cyclic Reduction (PCR) implementations by eliminating their hardware resource constraints. As a result, our parallel ADI, which is based on PCR, no longer has the maximum domain size limitation. Second, we optimize inefficient data accesses of parallel ADI solver by leveraging hardware texture memory and matrix transpose techniques. These memory optimizations further make already parallelized ADI solver twice faster, achieving overall more than 100 times speedup over a highly optimized CPU version. We also present the analysis of numerical accuracy of the proposed parallel ADI solver.

Composite Structured Mesh Generation With Automatic Domain Decomposition In Complex Geometries

Abstract This paper presents a novel automatic domain decomposition method to generate quality composite structured meshes in complex domains with arbitrary shapes, in which quality structured mesh generation still remains a challenge. The proposed decomposition algorithm is based on the analysis of an initial Delaunay triangulation on the closed boundary of the computational domain. The virtual edges with the so-called shortest effective length of the effective concave points are identified as candidates of the interface line between neighboring blocks. As demonstrated by examples and application, the proposed algorithm is capable of effectively decomposing complex domains without holes or islands (inner boundaries) into simpler patched blocks and thus significantly alleviated the difficulties of structured mesh generation in those domains.

Simulation of Storm Surge in the Mississippi Gulf Coast Using an Integrated Coastal Processes Model

This study uses an integrated coastal processes model to simulate hydrodynamics driven by storms, tides, river inflows, and winds in a large-scale domain covering the Mississippi and Louisiana Gulf Coasts. By using existing bathymetric data and DEM topographic data, a high-resolution mesh is generated to represent structures, roads, rivers, barrier islands, and lakes. For study of flooding and inundation of hurricanes in the inland areas of the Mississippi Gulf Coast, the Pearl River and its floodplain are included in the mesh with a detailed river course. Using two synthetic storms, four hurricane scenarios are simulated by using this model. Computed maximum storm surges without the Pearl River are compared with those with the river inflow. Differences between the storm surges indicate that the inclusion of the river inflow is imperative in order to obtain accurate predictions on flood and inundation due to storm surges in the Mississippi Coast community.

Boundary treatment for 2D elliptic mesh generation in complex geometries

This paper presents a boundary treatment method for 2D elliptic mesh generation in complex geometries. Corresponding to Neumann-Dirichlet boundary conditions (sliding boundary conditions), the proposed method aims at achieving orthogonal and smooth nodal distribution along irregular boundaries. In this method, a three-lined auxiliary mesh is constructed which is composed of the boundary line, an auxiliary line generated on its inner side, and the reflection line of this auxiliary line. The movements of the boundary nodes are determined by solving this auxiliary mesh. The boundary nodes are further corrected and adjusted by a second-order parabolic interpolation method and a weighting parameter depending on the curvature of the boundary. The proposed method was demonstrated through examples, and applied to field cases with complex geometries. It has been shown that this method is stable and effective in producing high quality meshes.

Visualization of Urban Area Flood Simulation in Realistic 3D Environment

A realistic 3D terrain was generated for part of the areas flooded due to 17th Street, New Orleans levee breaching during Hurricane Katrina by digital image processing techniques and GIS using LIDAR data in conjunction with high resolution multispectral satellite imagery. The DEM was used to generate mesh for simulating flood by CCHE2D Flood models. Propagation of the simulated flood was successfully visualized in realistic 3D environment by image draping technique using high resolution color orthophotographs. Obtained results indicate that (1) high resolution satellite imagery can be very useful to treat LIDAR data for realistic 3D terrain data generation, (2) ortho-rectified high resolution multispectral satellite data or color aerial photographs can be used in conjunction with the treated LIDAR data to generate true color realistic 3D terrain, and (3) propagation of simulated flood generated by CCHE2D Flood models can be visualized in realistic 3D environment using ArcGIS ArcScene software package.

Numerical modeling of flow through a breached levee and during levee closure.

Excessive precipitation, fast snow melting, and storm surges due to hurricanes often result in levee breaching and inland flooding which devastates social facilities, residential properties, environment quality and threatens human lives. To enhance the resilience of society, it is important and necessary to study levee breaching flows and develop technologies for repairing damaged levees effectively during the flood. In this study, computational simulations were conducted using a finite element based two-dimensional and depth-integrated numerical model, CCHE2D. Data of a scaled levee breaching physical model were used to validate the computational model. The flood and levee breaching occurred during the Hurricane Katrina in New Orleans, LA, were studied in the Hydraulic Lab, and the measured flow distribution in the vicinity of the breached levee was simulated. The numerical model resulted in good agreement with the measured flow velocity and water surface elevation distribution. The computational model such validated was also applied to simulate the breach closure processes conducted in the physical model. This capability could help emergency managers and engineers to close breached levees more effectively. The practice of placing sand bags into the flow and closing the breaching was mimicked by using a graphic user interface.

Edge gradients evaluation for 2D hybrid finite volume method model

ABSTRACT This paper presents a two-dimensional depth-integrated hydrodynamic flow model using the Finite Volume Method (FVM) on a hybrid unstructured mesh system with a collocated variable arrangement. A new evaluation method for computing the gradients at the cell edges, derived based on the second-order Taylor series expansion, is proposed to improve the solution due to mesh irregularity and non-uniformity. In this method, no interpolation for variables at cell vertices is necessary. The cross-edge flux is evaluated by using the momentum interpolation. The convergence, robustness and accuracy of this model based on the new evaluation method of edge gradients have been demonstrated by several numerical convergence tests and applications to natural rivers with complex geometries and instream structures.

Parallelization of Implicit CCHE2D Model using CUDA Programming Techniques

Applications of Computational Fluids Dynamics (CFD) analysis are strongly limited by computing efficiency particularly for large-scaled and long-term problems. In order to alleviate the efficiency problem, parallel computing, decomposing a large problem into multiple small problems, and solving them concurrently or simultaneously (in parallel), is increasingly adopted. In recent years, Graphics Processing Units, GPU, and particularly General-Purpose computation on Graphics Processing Units, GPGPU, have been successfully used for parallel computing in the areas of medical imaging, environmental science, and CFD, etc. The availability of a large number of processors greatly speeds up the efficiency of parallel computing. This study is aimed at parallelizing the implicit CCHE2D model (Jia et al., 2002). CCHE2D is a general depth integrated hydrodynamic model with sediment transport, water quality evaluation, chemical spill, and flood modeling capabilities. The flow model of CCHE2D is paralleled on GPU with CUDA Fortran programming techniques to improve its computational efficiency in personal computers (PC). Comparisons of numerical results and computing efficiency between the original sequential model and the parallelized model have shown that the parallelized CCHE2D model is consistent with the original model in producing reasonable results with significantly higher efficiency.

An improved nearly-orthogonal structured mesh generation system with smoothness control functions

This paper presents an improved nearly-orthogonal structured mesh generation system with a set of smoothness control functions, which were derived based on the ratio between the Jacobian of the transformation matrix and the Jacobian of the metric tensor. The proposed smoothness control functions are capable of relaxing the local strong orthogonal conditions so that nearly orthogonal but smooth mesh can be achieved. Examples and applications are also investigated in this paper to demonstrate the effects of the proposed mesh generation system.