Fernando Camelli

Running Unstructured Grid Based CFD Solvers on Modern Graphics Hardware

Techniques used to implement an unstructured grid solver on modern graphics hardware are described. The three-dimensional Euler equations for inviscid, compressible flow are considered. Effective memory bandwidth is improved by reducing total global memory access and overlapping redundant computation, as well as using an appropriate numbering scheme and data layout. The applicability of per-block shared memory is also considered. The performance of the solver is demonstrated on two benchmark cases: a missile and the NACA0012 wing. For a variety of mesh sizes, an average speed-up factor of roughly 9.5x is observed over the equivalent parallelized OpenMP-code running on a quad-core CPU, and roughly 33x over the equivalent code running in serial.

A linelet preconditioner for incompressible flow solvers

A parallel linelet preconditioner has been implemented to accelerate finite element (FE) solvers for incompressible flows when highly anisotropic meshes are used. The convergence of the standard preconditioned conjugate gradient (PCG) solver that is commonly used to solve the discrete pressure equations, greatly deteriorates due to the presence of highly distorted elements, which are of mandatory use for high Reynolds‐number flows. The linelet preconditioner notably accelerates the convergence rate of the PCG solver in such situations, saving an important amount of CPU time. Unlike other more sophisticated preconditioners, parallelization of the linelet preconditioner is almost straighforward. Numerical examples and some comparisons with other preconditioners are presented to demonstrate the performance of the proposed preconditioner.

Optimal placement of sensors for contaminant detection based on detailed 3D CFD simulations

Purpose – Develop a method for the optimal placement of sensors in order to detect the largest number of contaminant release scenarios with the minimum amount of sensors.Design/methodology/approach – The method considers the general sensor placement problem. Assuming a given number of sensors, every release scenario leads to a sensor input. The data recorded from all the possible release scenarios at all possible sensor locations allow the identification of the best or optimal sensor locations. Clearly, if only one sensor is to be placed, it should be at the location that recorded the highest number of releases. This argument can be used recursively by removing from further consideration all releases already recorded by sensors previously placed.Findings – The method developed works well. Examples showing the effect of different wind conditions and release locations demonstrate the effectiveness of the procedure.Practical implications – The method can be used to design sensor systems for cities, subway st...

Assessing maximum possible damage for contaminant release events

The combined use of damage criteria, genetic algorithms and advanced CFD solvers provides an effective strategy to identify locations of releases that produce maximum damage. The implementation is simple and does not require any change to flow solvers. A rather general criterion has been formulated to determine the damage inflicted by the intentional or unintentional release of contaminants. Results of two typical cases show that damage can vary considerably as a function of release location, implying that genetic algorithms are perhaps the only techniques suited for this type of optimization problem.

VLES Study of MUST Experiment

The focus of the current paper is the Mock Urban Setting Test (MUST) carried out at Dugway Proving Ground. The MUST experiment was designed to represent an urban complex of about 120 buildings with symmetric characteristics. Throughout the 19 day MUST experiment, 63 pu and continuous releases were carried out using propylene as a tracer gas. In order to understand the importance of capturing these spatial variations and unsteadiness, the multipurpose nite element code FEFLO-URBAN was used to perform a Very Large Eddy Simulation (VLES) of MUST. One of the continuous release trials (2682353) was selected for a detailed study. The terrain surface was modeled with geometric roughness in order to circumvent the problem of lack of turbulence production in the vertical direction. The FEFLO-URBAN simulations of the concentrations of the passive tracer were compared with the experimental measurements, resulting in agreement within an order of magnitude about 76% of the time. A study of the sensitivity of the model results with respect to resolution was made with four dieren t mesh resolutions (500K, 4M, 8.1M and 32M elements). A sensitivity study with the wind direction ino w boundary condition was carried out.

Porting of an Edge-Based CFD Solver to GPUs

Graphics processing units (GPUs) are increasingly becoming a mainstream platform for high performance computational fluid dynamics. This paper describes the porting of a substantial portion of FEFLO, an adaptive, edge-based finite element code for the solution of compressible and incompressible flow, to run on GPUs. The code is primarily written in Fortran 77 and has been ported to vector, shared memory parallel (via OpenMP) and distributed memory parallel (via MPI) machines. Due to the large scale of FEFLO and the likelihood of human error in porting, specialized code was written to perform automated translation from the OpenMP-parallelized edge and point loops to GPU kernels, implemented in either CUDA C or CUDA Fortran.

An ArcScene plug-in for volumetric data conversion, modeling and spatial analysis

Representation of volumetric data is necessary in hydrology, meteorology, ecology, environmental science and so on for modeling and visualization. As popular GIS software, ArcScene supports MultiPatch format for modeling and spatial analysis of volumetric data. It also supports conversion of 3D file formats into MultiPatch. But there are obvious limitations: (1) ArcScene does not support conversion of popular 3D formats such as Wavefront Objects (OBJ) and Visualization ToolKit (VTK) into MultiPatch; (2) after the conversion, the entire dataset only forms one patch which is inconvenient for further spatial analysis; (3) ArcScene does not support 3D contouring. In this study, we have successfully developed an ArcScene plug-in for converting OBJ and VTK file formats into MultiPatch, for accessing and editing MultiPatch files, and for tracing 3D contours based on MultiPatch. The ArcScene plug-in is characteristic of a time complexity of O(N), strong data structure in terms of spatial analysis, user-friendly operation, high robustness and good extendibility, and therefore it enriches ArcScene functionalities and users can benefit from the new plug-in for various applications. An Arcscene plug-in for volumetric data modeling based on MultiPatch: (a) urban pollution; (b) patches construction; (c) overhang; (d) overpass; (e) 3D holes; and (f) gasplumes. Display Omitted We have successfully developed an ArcScene plug-in.Converting volumetric data in OBJ/VTK file format into Multipatch.Tracing 3D contours based on Multipatch.We prove the efficiency of our ArcScene plug-in using Oklahoma City data.

Generating seamless surfaces for transport and dispersion modeling in GIS

A standard use of triangulation in GIS is to model terrain surface using TIN. In many simulation models of physical phenomena, triangulation is often used to depict the entire spatial domain, which may include buildings, landmarks and other surface objects in addition to the terrain surface. Creating a seamless surface of complex building structures together with the terrain is challenging and existing approaches are laborious, time-consuming and error-prone. We propose an efficient and robust procedure using computational geometry techniques to derive triangulated building surfaces from 2D polygon data with a height attribute. We also propose a new method to merge the resultant building surfaces with the triangulated terrain surface to produce a seamless surface for the entire study area. Using Oklahoma City data, we demonstrate the proposed method. The resultant surface is used as the input data for a simulated transport and dispersion event in Oklahoma City. The proposed method can produce the seamless surface data to be used for various types of physical models in a fraction of the time required by previous methods.

Integrating computational fluid dynamics (CFD) models with GIS: An evaluation on data conversion formats

Computational fluid dynamics (CFD) models are powerful computational tools to simulate urban-landscape scale atmospheric dispersion events. They are proven to be very useful for security management and emergency response. Essential inputs to CFD models include landscape characteristics, which are often captured by various GIS data layers. While it is logical to couple GIS and CFD models to take advantage of available GIS data and the visualization and cartographic rendering capabilities of GIS, the integration of the two tools have been minimal. In this paper, we took the first step to evaluate the use of GIS data in CFD modeling. Specifically, we explore how efficient is to use GIS data in CFD models and how sensitive the CFD results are to different GIS data formats. Using campus topography and building data, and the FEFLO-URBAN CFD model, we performed atmospheric release simulations using topographic data in contour and raster formats. We found that using raster format was quite efficient and contour data required significant effort. Though the simulation outputs from the two data formats were not identical, their overall outcomes were similar and did not post alarming discrepancies. We concluded that using GIS data have tremendous potential for CFD modeling.

Timings of an Unstructured-Grid CFD Code on Common Hardware Platforms and Compilers

Timings have been conducted for several benchmark testcases using the same code on a variety of common hardware platforms and compilers. The results indicate surprisingly little variation in the performance given the considerable number of vendors, architectures and compilers. The largest differences amounted to less than 30% of run-times. For the single processor runs, an increase in cache size reduced run-times, though not dramatically (approximately 10%). Going from 32 bits to 64 bits (with the same clockspeed, cache size, memory and compiler) in most cases produced a gain of 10%, although in some cases no gain was recorded. Overall, the chip with the slowest clocktime (Intel It II at 1.50 GHz) achieved the best performance. In some cases, it beat the 64-bit, 3.40 GHz P4/Xeon machines by a considerable margin. For shared memory parallelism, the best scaling was achieved by the SGI Altix. This should not come as a surprise, as SGI’s CCNUMA technology has matured over the last decade. For the AMD Opteron, the SUN compiler exhibited the best scaling.