Unstructured mesh generation based on Parallel Virtual Machine in cyber-physical system

A new physically-based approach to unstructured mesh generation via Monte-Carlo simulation is proposed. Geometrical objects to be meshed are represented by systems of interacting particles with a given interaction potential. A new way of distributing nodes in complex domains is proposed based on a concept of dynamic equilibrium ensemble, which represents a liquid state of matter. The algorithm is simple, numerically stable and produces uniform node distributions in domains of complex geometries and different dimensions. Well-shaped triangles or tetrahedra can be created by connecting a set of uniformly-spaced nodes. The proposed method has many advantages and potential applications.;The new method is applied to the problem of meshing of voxel-based objects. By customizing system potential energy function to reflect surface features, particles can be distributed into desired locations, such as sharp corners and edges. Feature-preserved surface mesh can then be constructed by connecting the node set.;A heuristic algorithm using an advancing front approach is proposed to generate triangulated surface meshes on voxel-based objects. The resultant surface meshes do not inherit the anisotropy of the underlying hexagonal grid. However, the important surface features, such as edges and corners may not be preserved in the mesh.;To overcome this problem, surface features such as edges, corners need to be detected. A new approach of edge capturing is proposed and demonstrated. The approach is based on a Laplace solver with incomplete Jacobi iterations, and as such is very simple and efficient. This edge capturing approach combined with the mesh generation methods above forms a simple and robust technique of unstructured mesh generation on voxel-based objects.;A graphical user interface (GUI) capable of complex geometric design and remote simulation control was implemented. The GUI was used in simulations of large fuel-cell stacks. It enables one to setup, run and monitor simulations remotely through secure shell (SSH2) connections. A voxel-based 3D geometrical modeling module is built along with the GUI. The flexibility of voxel-based geometry representation enables one to use this technique for both geometric design and visualization of volume data.

The linear quadtree is a useful data structure for representing an image for the sake of the storage saving and further image manipulations. In this paper we propose a linear quadtree coding scheme and implement this algorithm on the parallel virtual machine (PVM). Our goal is to demonstrate the applicability of using the PVM in combining the computing power of computers in a network to solve this kind of image processing problems. The processors in the PVM are organized as a master-slave paradigm and various numbers of processors are applied for different PVM’s to compare their performances. Experimental results show that the speedup of solving this image encoding problem in parallel is quite satisfactory. With such a PVM environment which is easily accessible in the public domain, high performance computing is truly possible without additional hardware cost.

A utility-maximizing tasks assignment method for rendering cluster system based on feedback is proposed to solve the problem that traditional task-centered assignment method and naive load balancing strategy cannot make full use of resources. The method uses feedback on resource usage to choose an appropriate number of threads for the renderer and then divides computing nodes of the rendering cluster system into fine grain computing units. After that, the method takes advantage of frame-to-frame coherence to assign tasks to computing units with a new static load balancing strategy. Experiments on two scene models with different complexity and comparisons with the naive method and the fixed-threads method show that the utility-maximizing method not only reduces the rendering time of a rendering job, but also balances the load between computing units. The scalability of the proposed method is also verified on different number of computing nodes.

In this paper we present our efforts to parallelize an unstructured quadrilateral mesh generator. Its serial version is based on the divider-and-conquer idea, and mainly includes two stages, i.e. geometry decomposition and mesh generation. Both stages are parallelized separately. A couple of parallel models are introduced and compared to parallelize the stage of geometry decomposition. A fine-grain level parallel algorithm proves preferable to that based on the task-dependency tree, with which the load imbalance brought by the improper utilization of the symmetry of the vertex pair matrix is removed nicely. Since the number of elements in sub-domains could be pre-computed before meshing, a simple static load balancing scheme is investigated, and the effect of granularity is also discussed briefly. Finally, experiments are designed to evaluate the performance of the parallel mesh generator in detail.

Traditional 3D graphics systems, e.g. Denali, with texture mapping capability implement the standard graphics pipeline in a pipelined parallel architecture. The systems are sufficiently powerful and flexible to support efficiently applications such as volume rendering. We discuss how Denali, manufactured by Kubota Graphics Corporation accelerates volume rendering. Volume rendering methods including maximum intensity projection and iso-surface rendering are implemented using Denali's 3D texture mapping capability. The volume rendering realization utilizes a partitioning data allocation scheme instead of a replicating approach to reduce dramatically texture memory requirements. Denali has two major parallel components: Transformation and Rasterization Modules (TRMs) and Frame Buffer Modules (FBMs). The flexible TRMs, containing a general purpose RISC processor, use object parallelism. The FBMs, containing hardware ASICs and pixel memory, use pixel parallelism. The TRMs and FBMs use dynamic and static load balancing schemes respectively to assign a roughly equal amount of computation to each parallel node. >

Generation of block-structured grids in complex computational domains using templates

Multi-objective unstructured triangular mesh generation for use in hydrological and land surface models

The volume of fluid (VoF) method is widely used in multi-phase flow simulations to track and locate the interface between two immiscible fluids. A major bottleneck of the VoF method is the interface reconstruction step due to its high computational cost and low accuracy on unstructured grids. We propose a machine learning enhanced VoF method based on Graph Neural Networks (GNN) to accelerate the interface reconstruction on general unstructured meshes. We first develop a methodology to generate a synthetic dataset based on paraboloid surfaces discretized on unstructured meshes. We then train a GNN based model and perform generalization tests. Our results demonstrate the efficiency of a GNN based approach for interface reconstruction in multi-phase flow simulations in the industrial context.

A robust Delaunay-AFT based parallel method for the generation of large-scale fully constrained meshes

To meet the objective of minimizing the job execution time, parallel computing has to deal with a lot of issues which crop up while working with parallel code. These issues can result in bottleneck and restrict the behavior of parallel program in attaining an aforesaid speedup as suggested by Amdahl Gene. The most problematic issue that crops up is the distribution of workload in both the categories of parallel system viz. homogeneous and heterogeneous systems. This situation demands an effective load balancing strategy to be in place in order to ensure a uniform distribution of load across the board. The scheduling `m' jobs to `n' resources with the objective to optimize the QoS parameters while balancing the load has been proven to be NP-hard problem. Therefore, a heuristic approach can be used to design an effective load balancing strategy. In this paper, a centralized dynamic load balancing strategy using adaptive thresholds has been proposed for a parallel system consisting of multiprocessors. The scheduler continuously monitors the load on the system and takes corrective measures as the load changes. The threshold values considered are adaptive in nature and are readjusted to suite the changing load on the system. Therefore, the scheduler always ensures a uniform distribution of the load on the processing elements with dynamic load environment.

The limiting factor in simulating flows past realistic configurations of interest has been the discretization of the physical domain on which the governing equations of fluid flow may be solved. In an attempt to circumvent this problem, many Computational Fluid Dynamic (CFD) methodologies that are based on different grid generation and domain decomposition techniques have been developed. However, due to the costs involved and expertise required, very few comparative studies between these methods have been performed. In the present work, the two CFD methodologies which show the most promise for treating complex three-dimensional configurations as well as unsteady moving boundary problems are evaluated. These are namely the structured-overlapped and the unstructured grid schemes. Both methods use a cell centered, finite volume, upwind approach. The structured-overlapped algorithm uses an approximately factored, alternating direction implicit scheme to perform the time integration, whereas, the unstructured algorithm uses an explicit Runge-Kutta method. To examine the accuracy, efficiency, and limitations of each scheme, they are applied to the same steady complex multicomponent configurations and unsteady moving boundary problems. The steady complex cases consist of computing the subsonic flow about a two-dimensional high-lift multielement airfoil and the transonic flow about a three-dimensional wing/pylon/finned store assembly. The unsteady moving boundary problems are a forced pitching oscillation of an airfoil in a transonic freestream and a two-dimensional, subsonic airfoil/store separation sequence. Accuracy was accessed through the comparison of computed and experimentally measured pressure coefficient data on several of the wing/pylon/finned store assembly's components and at numerous angles-of-attack for the pitching airfoil. From this study, it was found that both the structured-overlapped and the unstructured grid schemes yielded flow solutions of comparable accuracy for these simulations. This study also indicated that, overall, the structured-overlapped scheme was slightly more CPU efficient than the unstructured approach.

Unstructured mesh generation from the Virtual Family models for whole body biomedical simulations

A new method of unstructured mesh generation is presented. The advancing front method and Delauney triangulation have been commonly applied to it. In the former method, interior mesh nodes are generated based on the nodes on boundaries and parameters defined by the user. Unsuitable setting of parameters, however, may lead to failure in mesh generation. On the other hand, the latter method can accomplish mesh generation without failure, though the user has to set all the nodes. The present method, based on Delauney triangulation, requires the user to set only a few nodes on boundaries, and the other mesh nodes are generated automatically according to the curvature or proximity of boundaries. The perfomance of the presented method is illustrated in a region with multiple bodies.

Feng, X.; Wan, M., and Li, S., 2020. The design of marine remote sensing image fusion model based on dynamic load balancing strategy. In: Yang, Y.; Mi, C.; Zhao, L., and Lam, S. (eds.), Global Topics and New Trends in Coastal Research: Port, Coastal and Ocean Engineering. Journal of Coastal Research, Special Issue No. 103, pp. 943–946. Coconut Creek (Florida), ISSN 0749-0208.In this work, a model of marine remote sensing image fusion based on dynamic load balancing is proposed. The remote sensing image acquisition system based on Android intelligent device obtains the marine remote sensing image, and enhances and registers the image based on the fuzzy theory and template matching, respectively, to realize the image preprocessing. Based on the preprocessing results, considering the small number of objects in the marine remote sensing image, the target area is divided by the idea of quad tree structure. At the same time, the task allocation strategy of dynamic load balancing and the idea of parallel computing are used to propose the task allocation algorithm processing model of the cluster system for the fusion processing of the image in the target area, so as to achieve the fusion of marine remote sensing image. It can be known that the proposed model takes less time and has superior average gradient and clarity of image.

Development of boundary aligned grid generation is presented together with comparative performance of cell-vertex versus cell-centred CVD-MPFA finite-volume formulations using equivalent degrees of freedom. When generating structured or unstructured grids for reservoir simulation, classical key constraints involve boundary aligned grid generation with control-volume boundaries aligned with solid walls and geological features such as layers, shale barriers, fractures, faults, pinchouts and multilateral wells. The schemes used are control-volume distributed (CVD) with flow variables and rock properties sharing the same control-volume location and are comprised of a multipoint flux family formulation (CVD-MPFA). Consequently a natural choice is for primal grid cells to act as control-volumes, then grid generation can be performed with primal grid cell boundaries being aligned with key interior constraint boundaries. This naturally leads to cell-centred approximation, where flow variables and rock properties are associated with grid cell centres. The alternative is to employ cell-vertex approximation which uses far fewer approximation points on a given unstructured grid. In this case control-volumes are constructed around primal grid vertices. The grid generation process is less straight forward since control-volumes must be constrained to satisfy interior boundary alignment. A novel grid generation procedure is proposed that automates control-volume boundary alignment and yields a Voronoi mesh. The actual grid is then generated such that dual boundaries are aligned with key internal constraint boundaries and cell-vertex approximation becomes the natural choice. In this case flow variables and rock properties are associated with grid cell vertices and their dual control-volumes. The relative benefits of both types of approximation is made clear in terms of flow resolution and degrees of freedom required.