Massively parallel mesh generation for physics codes

Distributed parallel Delaunay mesh generation

Parallel and distributed adaptive quadrilateral mesh generation

A bottom-up approach to parallel anisotropic mesh generation is presented by building a mesh generator starting from the basic operations of vertex insertion and Delaunay triangles. Applications focusing on high-lift design or dynamic stall, or numerical methods and modeling test cases, still focus on two-dimensional domains. This automated parallel mesh generation approach can generate high-fidelity unstructured meshes with anisotropic boundary layers for use in the computational fluid dynamics field. The anisotropy requirement adds a level of complexity to a parallel meshing algorithm by making computation depend on the local alignment of elements, which in turn is dictated by geometric boundaries and the density functions— one-dimensional spacing functions generated from an exponential distribution. This approach yields computational savings in mesh generation and flow solution through well-shaped anisotropic triangles instead of isotropic triangles. The validity of the meshes is shown through solution characteristic comparisons to verified reference solutions. A 79% parallel weak scaling efficiency on 1,024 distributed memory nodes, and a 72% parallel efficiency over the fastest sequential isotropic mesh generator on 512 distributed memory nodes, is shown through numerical experiments.

A bottom-up approach to parallel anisotropic mesh generation is presented by building a mesh generator from the principles of point-insertion, triangulation, and Delaunay refinement. Applications focusing on high-lift design or dynamic stall, or numerical methods and modeling test cases use two-dimensional domains. Our push-button parallel mesh generation approach, meaning the user only needs to start the program by specifying the initial geometry, anisotropic gradation, and ray angle constraint, can generate high-fidelity unstructured meshes with anisotropic boundary layers for use in the computational fluid dynamics field. © 2015 The Authors. Published by Elsevier Ltd. Peer-review under responsibility of organizing committee of the 24th International Meshing Roundtable (IMR24).

This paper describes our progress in creating a case study on optimistic computing for the Grid using parallel mesh generation. For the implementation of both methods we use a portable runtime environment for mobile applications (PREMA) which is extended to provide support for optimistic control using grid performance monitoring and prediction. Based on the observed performance of a world-wide grid testbed, we use this case study to develop a methodology for estimating target operating regions for grid applications. The goal of this project is to generalize the experience and knowledge of optimistic grid computing gained through mesh generation into a tool that can be applied to tightly coupled computations in other application domains.

Creating in parallel guaranteed quality large unstructured meshes is a challenging problem. Parallel mesh generation procedures decompose the original mesh generation problem into smaller subproblems that can be solved in parallel. The subproblems can be treated as either completely or partially coupled, or they can be treated as completely decoupled. In this paper we present a parallel guaranteed quality Delaunay method for 2-dimensional domains which is based on the complete decoupling of the subproblems. As a result the method eliminates the communication and the synchronization during the meshing of the subproblems. Moreover, it achieves 100% code reuse of existing, fine-tuned, and well-tested sequential mesh generators. The approach we describe in this paper presents for the first time an effective way to create in parallel guaranteed quality meshes with billions of elements in a few hundreds of seconds, and at the same time demonstrates that these meshes can be generated in an efficient and scalable way. Our performance data indicate superlinear speedups.

Parallel adaptive tetrahedral mesh generation by the advancing front technique

This paper describes a parallel approach for the rapid generation of ultra-large-scale unstructured meshes on distributed-memory supercomputers. A medium-sized initial mesh is prepared first. Afterwards, a two-level domain decomposition (DD) strategy is used to split and distribute the initial mesh to different cores. Finally, the parallel mesh generation, comprising a recursive procedure which includes parallel surface recovery, parallel boundary updating, and parallel mesh multiplication, is performed. The two-level DD differentiates the intra-node and inter-node communication to reduce communication overheads. A global indexing and updating scheme is used to make the mesh multiplication devoid of communication. A new parallel surface recovery algorithm without communication is developed to maintain the fidelity of the resulting mesh model to the original geometric model. Tests of the parallel approach for some real-life problems on supercomputers (Dawning-5000A and Tianhe-2) are presented. Issues regarding the speedup, parallel efficiency, and mesh quality are discussed. Results show that the proposed parallel approach has a reasonably good scalability, that the quality of the resulting mesh is improved, and that ultra-large-scale meshes with billions of elements can be generated quickly.

A significant amount of time in finite element modeling of manufacturing processes is spent in mesh generation. Setting up three-dimensional meshes is a cumbersome task due to complexity of the processes and the involved geometries. Moreover, additional meshing challenges often appear due to the fact that manufacturing processes based on large plastic deformations present progressive mesh distortion (or degeneracy), potential interference between mesh and contour of the tools and possible contact of the mesh with itself. This poses the need for robust, automatic, mesh generation and regeneration (remeshing) procedures in order to ensure that complex processes are modeled from the beginning to the end with high levels of accuracy both in terms of geometry and distribution of field variables.

Simulation is a key component of surgical training, enabling trainees to develop their skills in a safe environment. With simulators broadly grouped into physical models and virtual-reality (VR) simulators, it is important to evaluate the comparative effectiveness of the simulator types in terms of validity as well as cost. The review aims to compare the benefits and drawbacks of novel VR and physical simulators within the broader themes of endourology, laparoscopic and robotic operations, and other urological procedures. Key benefits of bench models include their comparatively lower cost, easy access and provision of haptic feedback, whereas VR simulators are generally self-sufficient, reusable and enable skills of haemostasis to be practised. The advent of perfused 3D printed simulators across a range of urological procedures may replace cadavers as the traditional gold-standard simulation modality. Although possessing differing strengths and downsides, VR and physical simulators when used together can have an additive effect due to skill transferability across the platforms. Further comparative studies are required to directly quantify the differences between physical models and VR simulators in terms of performance metrics and cost-effectiveness. There is lack of validated VR simulators for open and reconstructive procedures.

Computational simulation must often be performed on domains where materials are represented as scalar quantities or volume fractions at cell centers of an octree-based grid. Common examples include bio-medical, geotechnical or shock physics calculations where interface boundaries are represented only as discrete statistical approximations. In this work, we introduce new methods for generating Lagrangian computational meshes from Eulerian-based data. We focus specifically on shock physics problems that are relevant to ASC codes such as CTH and Alegra. New procedures for generating all-hexahedral finite element meshes from volume fraction data are introduced. A new primal-contouring approach is introduced for defining a geometric domain. New methods for refinement, node smoothing, resolving non-manifold conditions and defining geometry are also introduced as well as an extension of the algorithm to handle tetrahedral meshes. We also describe new scalable MPI-based implementations of these procedures. We describe a new software module, Sculptor, which has been developed for use as an embedded component of CTH. We also describe its interface and its use within the mesh generation code, CUBIT. Several examples are shown to illustrate the capabilities of Sculptor.

Scalable 3D Hybrid Parallel Delaunay Image-to-Mesh Conversion Algorithm for Distributed Shared Memory Architectures

In this paper, we introduce a novel, SPH-based(Smoothed Particle Hydrodynamics), parallel adaptive quadrilateral-dominated mesh generation method. Building upon particle-based mesh generation techniques, we propose an innovative Adaptive Smoothing Length Smoothed Particle Hydrodynamics (ASL-SPH) Method utilizing the L∞ norm metric to enhance particle relaxation. The core concept leverages the square characteristic of the L∞ norm's unit circle to facilitate the construction of high-quality quadrilateral meshes. The algorithm comprises three primary stages: (1) computation of the initial adaptive particle distribution based on target size and density fields; (2) GPU-accelerated particle relaxation guided by a direction field, employing the proposed ASL-SPH grounded in the L∞ norm metric; and (3) establishment of topological connections among particles to generate a quadrilateral-dominated mesh. Numerical experiments validate that the proposed method effectively generates high-quality quadrilateral-dominated meshes under complex boundary conditions. Furthermore, the integration of GPU parallel computing significantly enhances the algorithm's efficiency.

A steam injection test, carried out in the 1.5 m physical simulator developed at the Alberta Research Council for the evaluation of in situ recovery processes for oil sands is described. The experimental facilities and procedures, together with the test data obtained are described. Numerical modeling studies performed to simulate the test run using an implicit steam model are outlined. The capabilities of the numerical model, its mathematical formulation are described briefly without using equations. Relevant data for the simulation are given and the experimental results are compared with the model predictions. Introduction Physical and mathematical models are used to obtain 1n understanding of the recovery processes of bitumen from oil sands by steam injection. Both these approaches, in principle, can be used to extrapolate laboratory data to the field and predict the performance of a reservoir. Physical modeling, using scaling factors obtained from dimensional and inspectional analysis, even though desirable, is in fact difficult because, not all relevent mechanisms can be scaled. Mathematical models, on the other hand can be used effectively provided they have been validated by experiments. These experiments can be performed on reservoir elements or elemental models1. The materials and operating conditions in these elemental models are such that they represent some point in the actual reservoir and the recovery process involves the same mechanisms that operate in the reservoir. The physical shape and side of the elements is determined by the constraints of the test equipment. In the following sections, the experimental equipment at the Alberta Research Council, to studythe in situ recovery of bitumen from oil sands by steam injection, is briefly described. This equipment, generally known as the 1.5 m physical simulator facility, can be used for scaled models as well as elemental models. The numerical modelswhich have been developed at the Alberta Research Council to simulate steam floods in the physical simulator are introduced. A steam injection testcarried out in the 1,5 m physical simulator and its numerical simulation using an implicit steam model are discussed. EXPERIMENTAL EQUIPMENT A schematic diagram is shown in Figures 1 and 2. The equipment consists of a high pressure vessel, which contains the oil sand bed, three injection systems (steam, solvent and gas) and equipment to handle the production fluids. Process control and data collection is handled by a FOX 2/10, Foxboro minicomputer. The Physical Simulator The physical simulator is designed for a maximum operating pressure of 11.4 MPa. It has an internal diameter of 1.5 m with an overall height of 2.5 m. The vessel is separated into two parts by a flange. The lower part which contains the oil sand is 1.5 m high. The upper part is pressurized with nitrogen gas to simulate the overburden pressure. The nitrogen gas is separated from the oil sand bed by means of a deformable carbon steel plate of thickness L2 AWG, which transfers the overburden pressure to the oil sand bed.

The Seismic Acoustics Laboratory (SAL) has as its principal research facility a physical modeling system. Experiments conducted with this system in the area of 3D seismic profiling have led to practical applications of physical modeling for both exploration and production geophysicists. The physical models, which are scanned in a large water tank, have been fabricated to represent various structural and stratigraphic traps. Across these models, the scanning source and receiver are independently controlled which allows for an infinite number of data acquisition schemes. To insure proper scaling of the model to the earth prototype, the large transducer source is focused to a point and the subsequent propagating wavelet is consistent enough to allow wavelet processing of the data. The resulting time sections from the physical models have a high enough signal-to-noise ratio to allow comparisons to both theoretical models and field seismic data. Time sections which profile in the modeling tank a simple anticline and also a more complicated faulted delta compare favorably with theoretical time sections. Using these two geological models, we describe various applications and pitfalls of modeling which will be of interest to interpreters working in problem areas. Introduction Practical applications of physical modeling have been discussed in the past by several authors. l, 2,3,4,7 This paper is a non-mathematical update and overview of how geophysicists can use physical modeling, along with theoretical modeling, as an interpretational tool. Because many seismic processing techniques are based on the scalar wave equation, it will be interesting to note the differences between theoretical and physical model data. These differences will provide us with insight as to when our processing and/or interpretation might be incorrect with reference to different geological structures and lithology. The Physical Modeling System The physical modeling system at SAL was based on the system described by Frenchl and exact details are given in SAL Annual Reports. 5,6 Working with the scale factors which are given in Figure 1, an 8x6x5 ft water tank was designed and built in 1977. The tank's large size insured that reflections from the sides, top or bottom would be outside the time window of signals reflecting from the physical model in the middle of the water tank. Resting on top of the tank are the computer-controlled positioning mounts (Figure 2), one for the source and the other for the receiver. These mounts control the xyz-coordinates of the source and receiver within a .01 tolerance (equivalent prototype 10 ft). Figure 3 is a top view of the tank with the source and receiver positioned over a geological model resting on thin wires. The models built to date have been approximately 16 in square and up to 4 in thick. Multi-layer models are fabricated by pouring RTV silicone rubber into a plaster cast and after the rubber hardens, the second layer with its different RTV properties is poured between a second plaster cast and the first RTV layer. Models with four-layers have been built in this manner.