Abstract:

Visualization of porous media is of great importance to several scientific fields, including the petroleum technology. The topic of this thesis arises from our collaborations with The Center for Integrated Operations in the Petroleum Industry. By being able to quickly analyze properties of porous rocks, they can get a better understanding of how to efficiently harvest oil since oil is typically held and stored within rock pores. The petroleum industry typically uses Computed Tomography (CT) technology to scan rock samples for their internal structures. The resulting data is loaded into a computer program that generates 3D models of the rocks describing the 3D nature of its’ internal structure. The scan data created from these scans will in most cases contain inaccuracies due to artifacts created while scanning.
  

In this thesis, we develop an application that interactively helps the user localize the rock and pores in the CT scan data, allowing the user to create an image with a more accurate representation of the pores. We use digital image processing techniques to do an initial localization of the elements in the scan. The artifacts are then reduced by allowing the user to drag and pull on the line-data specifying the pores. Our implementation then uses this new representation to construct a 3D volume image that can be used in geophysical applications, like Schlumberger Petrel, for further analysis and simulation.

Abstract:

Today's petroleum industry demand an ever increasing amount of computational resources. Seismic processing applications in use by these types of companies have generally been using large clusters of compute nodes, whose only computing resource has been the CPU. However, using Graphics Processing Units (GPU) for general purpose programming is these days becoming increasingly more popular in the high performance computing area. In 2007, NVIDIA corporation launched their framework for developing GPU utilizing computational algorithms, known as the Compute Unied Device Architecture (CUDA), a wide variety of research areas have adopted this framework for their algorithms. This thesis looks at the applicability of GPU techniques and CUDA for o-loading some of the computational workload in a seismic shot modeling application provided by StatoilHydro to modern GPUs.

This work builds on our recent project that looked at providing check- point restart for this MPI enabled shot modeling application. In this thesis, we demonstrate that the inherent data parallelism in the core nite-dierence computations also makes our application well suited for GPU acceleration. By using CUDA, we show that we could do an efficient port our application, and through further renements achieve signicant performance increases.

Benchmarks done on two dierent systems in the NTNU IDI (Department of Computer and Information Science) HPC-lab, are included. One system is a Intel Core2 Quad Q9550 @2.83GHz with 4GB of RAM and an NVIDIA GeForce GTX280 and NVIDIA Tesla C1060 GPU. Our second testbed was an Intel Core I7 Extreme (965 @3.20GHz) with 12GB of RAM hosting an NVIDIA Tesla S1070 (4X NVIDIA Tesla C1060). On this hardware, speedups up to a factor of 8-14.79 compared to the original sequential code are achieved, conrming the potential of GPU computing in applications similar to the one used in this thesis. 

Abstract:

 We consider three high-resolution schemes for computing shallow-water waves as described by the Saint-Venant system and discuss how to develop highly efficient implementations using graphical processing units (GPUs). The schemes are well-balanced for lake-atrest problems, handle dry states, and support linear friction models. The first two schemes handle dry states by switching variables in the reconstruction step, so that that bilinear reconstructions are computed using physical variables for small water depths and conserved variables elsewhere. In the third scheme, reconstructed slopes are modified in cells containing dry zones to ensure non-negative values at integration points. We discuss how single and double-precision arithmetics affect accuracy and efficiency, scalability and resource utilization for our implementations, and demonstrate that all three schemes map very well to current GPU hardware. We have also implemented direct and close-to-photo-realistic visualization of simulation results on the GPU, giving visual simulations with interactive speeds for reasonably-sized grids.

Abstract:

The development of shape repositories and 3D databases rises the need of online visualization of 3D objects. The main issue with the remote visualization of large meshes is the transfer latency of the geometric information. The remote viewer requires the transfer of all polygons before allowing object’s manipulation. To avoid this latency problem, an approach is to send several levels of details of the same object so that lighter versions can be displayed sooner and replaced with more detailed version later on. This strategy requires more bandwidth, implies abruptly changes in object aspect as the geometry refines as well as a non negligible precomputing time. Since the appearance of a 3D model is more influenced by its normal field than its geometry, we propose a framework in which the object’s LOD is replaced with a single simplified mesh with a LOD of appearance. By using Appearance Preserving Octree-Textures (APO), this appearance LOD is encoded in a unique texture, and the details are progressively downloaded when they are needed. Our APO-based framework achieves a nearly immediate object rendering while details are transmitted and smoothly added to the texture. Scenes keep a low geometry complexity while being displayed at interactive framerate with a maximum of visual details, leading to a better visual quality over bandwith ratio than pure geometric LOD schemes. Our implementation is platform independent, as it uses JOGL and runs on a simple web browser. Furthermore the framework doesn’t require processing on the server side during the client rendering. 

Abstract:

Typically, flow volumes are visualized by defining their boundary as iso-surface of a level set function. Grid-based level sets offer a good global representation but suffer from numerical diffusion of surface detail, whereas particle-based methods preserve details more accurately but introduce the problem of unequal global representation. The particle level set (PLS) method combines the advantages of both approaches by interchanging the information between the grid and the particles. Our work demonstrates that the PLS technique can be adapted to volumetric dye advection via streak volumes, and to the visualization by time surfaces and path volumes. We achieve this with a modified and extended PLS, including a model for dye injection. A new algorithmic interpretation of PLS is introduced to exploit the efficiency of the GPU, leading to interactive visualization. Finally, we demonstrate the high quality and usefulness of PLS flow visualization by providing quantitative results on volume preservation and by discussing typical applications of 3D flow visualization.