A Multi-Resolution Interpolation Scheme for Path-Line Based Lagrangian Flow Representations

Size: px
Start display at page:

Download "A Multi-Resolution Interpolation Scheme for Path-Line Based Lagrangian Flow Representations"

Transcription

1 Vision, Modeling, and Visualization (2014) J. Bender, A. Kuijper, T. von Landesberger, and P. Urban (Eds.) A Multi-Resolution Interpolation Scheme for Path-Line Based Lagrangian Flow Representations Paper ID: 1014 Abstract The computation of trajectories in vector fields is typically performed in an iterative fashion through numerical interpolation. If the field is given as a set of trajectories the Lagrangian representation this integration can be replaced by interpolation. We present a novel re-organization of such Lagrangian representations by subsampling a flow map into multiple levels of resolution, maintaining a bound over the amount of memory mapped by the file system. The base resolution facilitates a real-time, low memory cost, interactive environment for flow field exploration. Once an area of interest is selected, additional trajectories from other resolution levels are dynamically loaded in the vicinity of the desired feature, providing for a low memory footprint. Through path- and time-surface extraction, we show that interpolation outperforms traditional advection and that dynamic trajectory refinement leads to results whose accuracy converges to that of features extracted through numerical integration. 1. Introduction Flow fields are the output of some of the most widely used simulations to date. In the context of Computational Fluid Dynamics (CFD) simulations, they are used to represent material transport in a wide range of application fields such as airplane engineering or climate simulations. The ability to trace individual particles present in the flow domain allows the visualization of material transport during post-processing. Given the classic representation of flow fields as static meshes with fluid velocities given per time step, the so-called Eulerian representation, this tracing requires the computation of solutions to the underlying discretized differential equation known as numerical integration. Due to the unpredictability of particle movement, it is difficult to determine which sections of the vector field data will be needed as the particle is advected through the field. Solutions that load the entire mesh, or load data on demand, have to take special care not to exceed memory (large amounts of data) and time (large amounts of I/O operations) constraints [CGC 11]. Alternatively, flow fields can be represented in a Lagrangian way, where the discretization is no longer governed by a stationary mesh. This representation corresponds to a dense sampling of the flow map, which encodes the advection of massless particles under the influence of the vector field over a given time interval. Such data can be, for example, output of modern CFD simulation tools (cf. Smooth Particle Hydrodynamics [Mon88]), or the output of another post-processing technique such as Finite-Time Lyapunov Exponent (FTLE) computations [Hal01]. The availability of a full Lagrangian representation in the form of a set of dense particle trajectories opens up new possibilities for integration-based flow visualization. In order to trace an arbitrary particle through the flow field, one no longer has to perform numerical integration, but can interpolate within the given set of flow trajectories. The predictive nature of interpolation, combined with the spatial coherency properties of continuous flow, allows for the use of a local interpolation neighborhood to extract the trajectory over all time steps. This locality assures a bound on the information needed to perform interpolation, with the advantage of reducing I/O to only those trajectories needed for the interpolation neighborhood. We demonstrate how such a particle trajectory representation can be used to construct adaptive integral surface representations, and compare the results to traditional integrationbased surface generation with respect to matched accuracy, improved speed, and a reduced memory footprint. The advantages of a Lagrangian representation of fluid flow are further exemplified by a real-time, interactive data set exploration technique. By segmenting the flow map representation into multiple resolutions, the coarsest resolution allows for real-time surface reconstruction, while finer resolutions refine a specific surface with on-demand tracer loading. This technique gives users the ability to explore the entire flow data set for a desired surface, and once located, the surface is

2 refined to the best possible accuracy (defined by the total set of trajectories). Only those trajectories needed to reconstruct any given surface accurately are loaded in, thereby lowering the memory footprint of surface reconstruction. In summary, the main contributions of this work to the field of visualization are: A multi-resolution flow map organization for dense particle tracer sets. Efficient interpolation of arbitrarily seeded trajectories. Accurate and efficient extraction of integration-based flow geometry at multiple levels of detail, allowing real-time flow field exploration. Location specific I/O for a reduction in memory cost during integral surface representation. 2. Related Work Integral surfaces, as created from sets of trajectories or particle positions, are a tool capable of visualizing complex flow behaviors. Typically, the computation of underlying trajectories makes heavy use of numerical integration techniques such as Euler integration or higher-order adaptive Runge- Kutta Methods [CK90]. Recent surface generation methods [MLP 10] have the advantage of adaptive resolution and accuracy control [Hul92, GKT 08, KGJ09] but require high numbers of computationally expensive and iterative evaluation of vector fields. Additionally, surfaces with comparatively low accuracy and constrained adaptivity may also be computed on the GPU [BFTW09]. To increase computational efficiency, different approaches such as stream function generation with subsequent isosurface extraction [vw93, WJE00], or path-volume construction from integrated time-surfaces [OHBH13], or hierarchical construction of the integral curves [HSW11] have been taken in the past. Other flow structure construction techniques [SWS07] make direct use of sample trajectories as well, but do not aim at the creation of adaptive integral surface structures. These approaches do not, however, easily extend to 3D space + time. In similar contexts, other research [WT10] formulates time-varying integral surfaces as tangent curves in high-dimensional spaces [TS03], significantly increasing pre-computation and memory requirements. Memory reduction has been achieved in the form of vector field sub-sampling and compression [LRR00] with topology conservation but it is difficult to estimate the impact on trajectory advection. A class of visualization techniques that makes heavy use of pre-sampled vector fields is the field of Lagrangian flow analysis with the help of Finite-Time Lyapunov Exponent maps [Hal01]. Here, densely sampled flow fields are the basis for the approximation of flow stretching [SP07, GGTH07]. These approaches are suitable for direct execution on massively parallel machines due to locality of trajectory integration. To a lesser degree, this locality can assist curve integration in parallel. Massively parallel stream line integration covering seeding strategies has been implemented and analyzed [CGC 11], with a strong focus on I/O, citing computational hardship due to block loading of the vector field because of the inability to predict a particle s position over all time steps prior to advection. Focusing directly on pre-sampled vector fields, our method restructures trajectory information to allow for dynamic loading of pre-determined blocks, eliminating I/O due to the position unpredictability of particle advection. With I/O limited to desired areas of interest, we are therefore able to present a method for path and time surface extraction that has real-time, interactive character even for large data sets. 3. Vector Field Representations A vector field is a continuous, vector-valued function v over a four-dimensional (space-time) domain of definition Ω [0,T ] R 3 that maps a point (t,x) to a vector v(t,x) R 3. A vector field is called stationary if v does not depend on t, and is time-dependent otherwise. Eulerian The classic representation of a vector field as a stationary mesh that holds vector data at its nodes for every time step of a simulation is called Eulerian. In this representation, the spatial discretization of the flow domain is generally either static in time, or does not reflect fluid transport. Lagrangian In a Lagrangian representation, frequently used in SPH and other mesh-free methods, the flow field consists of a collection of dynamically moving flow particles (tracers). In this representation, the frame of reference is locally moving with the flow field, i.e., the spatial discretization of the flow field is attached to participating material in the fluid transport, following the trajectories of individual particles. A comparison of both views is shown in Figure 1. Flow Trajectories The trajectory of a massless particle is identified by a point (t 0,x 0 ) it contains, commonly called its seed point. All subsequent positions, as obtained through advection, then form a pathline or trajectory. Mathematically, trajectories are solutions to the differential equation d I(t) =v(t,i(t)) dt (1) I(t 0 )=x 0 (t 0,x 0 ) Ω (2) Like the vector fields over which they are defined, trajectories typically possess a straightforward connection to physical aspects of a given application: e.g., in the case of fluid flow, integral curves describe the trajectories of massless particles embedded in a flow. Arbitrary flow trajectories may be reconstructed from both Eulerian and Lagrangian representations of the flow field. The former representation requires the use of numerical integration schemes, whereas the

3 φ t Eulerian: v( ) = Lagrangian: φ t( ) = Figure 1: Eulerian vs. Lagrangian descriptions. The Eulerian view describes a flow field in terms of the instantaneous velocity v of particles, while the Lagrangian view considers the evolution of particle positions. latter performs interpolation within a set of given trajectories. The work presented investgates and utilizes Lagrangian flow representations for efficient flow feature extraction. 4. A System for Efficient Path Line Interpolation The accurate reconstruction of trajectories in flow fields is heavily influenced by two key factors: spatial resolution of the underlying data and appropriate choice of interpolation scheme. We develop a system that investigates and utilizes the strengths of Lagrangian flow representations towards efficient multi-resolution reconstruction of pathlines. System Outline The Lagrangian nature of our data allows us to design a system that efficiently reconstructs pathlines through interpolation. The gained efficiency allows for an interactive flow field exploration environment, instantly presenting initial results while progressively improving them over time. The initial feature reconstruction relies on pathlines interpolated from a low resolution sampling of the flow map. Once a feature of interest is located, additional tracers are loaded to improve reconstruction accuracy. By organizing the available Lagrangian data into varying levels of resolution, only relevant tracers are located and used by the reconstruction algorithm. This process may be repeated any number of times during data set exploration, loading trajectory information only when requested. The central steps of our algorithm are captured in a flow diagram in Figure 2 and are as follows. 1. Pre-Processing: The Lagrangian representation is processed and written to files as sub-sampled flow maps consisting of multiple levels of detail. This file representation is designed to allow efficient local and on-demand streaming of trajectory data, following a hierarchy based on grid resolution. 2. Base Resolution: The flow map with the lowest available resolution is loaded and kept in memory for fast trajectory exploration and interpolation. 3. Efficient Interpolation: a. Interactive Extraction: During exploration, a preview version of the desired path- or time-surface is extracted in real time from the coarse level flow map rep- Figure 2: The process starts with an interactive exploration of a coarse resolution flow map. Once a feature of interest is located, the algorithm enters the refinement pipeline, loading additional tracers from finer resolutions. At this time, additional time step information may be loaded for the selected tracers. Once the seeding points are moved from the refined location, the process begins anew. resentation. For this matter, individual pathlines are interpolated from the existing set of flow map tracers. b. Spatial Refinement: On demand or after sufficient idle time has passed, higher resolution trajectory data is streamed from the file system. The existing surface representation is refined by taking novel trajectory data into account during pathline interpolation. The following section describes these steps in detail. 5. System Implementation We begin with an in-depth description of the data representation created by the pre-processing step, with a look at how tracer information is organized on the hard drive Organizing the Lagrangian Flow Representation If a small number of tracers is sufficient for reconstructing arbitrary pathlines under given accuracy constraints, loading the millions of tracers within a full resolution Lagrangian representation is an unnecessary burden. Therefore, the given Lagrangian representation is first pre-processed to place the tracers into multiple levels of resolution of varying density. Then, each resolution is segmented with a regular grid to allow for fast spatial location. Essentially, the entire Lagrangian representation is sub-sampled into smaller, separate, more computationally manageable representations. Each level of resolution breaks the data set up into a number of square voxels, or cells, of varying size, resulting in uniform grids of increasing resolution. An arbitrary time step τ of the vector field is chosen to represent the entire set of tracers. This means that a tracer s 3D position at τ is used to locate its designated cell. Beginning from the coarsest resolution, tracers are binned according to their tendency for

4 divergence, calculated as their FTLE value. This helps prioritize flow structures and material lines that play a key role in capturing representative structures. Binning continues until all tracers are exhausted and all resolution levels are filled. Resolution Specification The Lagrangian representation is broken up into multiple levels of resolution based on the data set dimensions, DIM xyz and the number of trajectories given, N total. The first resolution level to be created is the coarsest, providing a general subsampling over the entire domain. Each dimension is sampled by 100 cells, defining the cell width, CW c. If the data set does not have equal dimensions, the longest axis is represented by 100 cells, setting the resolution s cell width. The number of points within each cell of the coarse resolution, n c, is chosen to be 10, ensuring enough points in one cell to form a neighborhood for interpolation. To limit the amount of information mapped by the I/O, a value M is chosen where M represents the maximum number of points that should be mapped by the file system during pathline refinement. Setting a limit to the amount of information mapped at once assures that all resolutions store the same maximum amount of tracers. The number of levels, L, is defined by: N L = total N c (3) M where the numerator holds the amount of trajectories left to be stored outside of the coarsest resolution (N c being the total number of tracers in the coarsest resolution). Next, the cell size of each resolution is determined as CW i = CW c i where i is the current resolution level. With cell size known, the algorithm now determines how many tracers will be binned into each cell subject to the constraint, n i M (DIM x /CW i ) (DIM y /CW i ) (DIM z /CW i ) where i is the current resolution. The values n i will be used to sort tracers into the various resolution levels. There are two main objectives behind using multiple levels of resolution. First, sectioning the tracers into different levels lightens the load of the file system as resolutions are read in by the algorithm. Second, the varying density between resolutions helps ensure that loaded tracers may be used in the interpolation neighborhood of more than one seeded particle if a group of particles is seeded near one another. Binning of Trajectories For the purpose of locating a tracer in 3D, the input set of trajectories must be binned into the defined multi-resolution grids. Binning begins from the coarsest resolution to the finest, with the 3D position of a tracer at time τ used to determine which cell of the current resolution contains the tracer. A list of contained tracers is created for each cell, from which a subset is then selected. However, simply binning the n number of tracers can lead to an unbal- (4) Figure 3: Resolutions file layout. Resolutions are written in order with each resolution writing out its cells by collapsing the 3D representation into a linear one. anced cell. Therefore, a trajectory s spatial orientation determines whether or not it is chosen. Each cell has a designated radius r = 3 3 CW 3 i 8 n i that is used for spacing binned tracers, where CW i 3 n i is the cell dimensions divided by the number of trajectories to be binned. An empirically derived factor of 3 8 is applied to calculate the radius within the allotted space. If the list is exhausted before the n c criteria is met, trajectories furthest from those already selected are chosen to fill the cell. Any remaining tracers continue to the next resolution, where they are once again sorted into empty cells. Generating Resolution-Dependent Trajectory Files At the conclusion of the binning process, cells within all levels of resolution contain indices of their enclosed tracers. We write this data to files following resolution and cell order, one file for each time step. The order in which the trajectories appear in the file is based on their resolution and cell placement. Resolution levels are written in order from coarse to fine, while cells are laid out in the traditional xyz fashion with the x axis running fastest and the z axis slowest, shown in Figure 3. Trajectory information for the current time step is written out strictly according to this layout, a size of 16 bytes per tracer, or one 32-bit integer and 3 floats. The integer holds the index of the tracer and the 3 floats hold its 3D position. An additional integer holds the number of tracers within the current cell. The memory layout is shown in Figure 4. The first file to be written is the time step τ. Recall that all tracers were binned into the resolution cells using their spatial position at time τ. Even though a tracer s 3D position may be different at a future time step, the tracer s location within any time step file remains static, as specified by the binning process. It is possible that a tracer with a valid position at τ may have ceased to exist during the fluid simulation Figure 4: Cell contents within a resolution for one time step file. Each cell begins with an integer which holds the number of trajectories within the cell, after which each trajectory is represented by its index and 3D position.

5 and therefore, may not have position information for an arbitrary future time step t i. To preserve the uniform layout between time step files, dummy values are inserted to make up for the missing 16 bytes of information. With prior knowledge of initial tracer location, specific tracers can quickly be found in these files with one seek operation for the enclosing cell. This knowledge comes from a reference file that is output simultaneously with the first time step file τ. The reference file is structured similarly to the time step files, ordered by resolution and cell number. Unlike the other files, the reference file holds the number of bytes to be read into a time step file to locate a given cell. This allows for the flexibility of reading and storing only specified tracers, reducing the overall memory footprint and I/O computation time. The aforementioned process is part of a pre-processing step that does not run in real-time due to intense data sorting and writing to disk. Generating the complete set of resolution-dependent trajectory files takes approximately 8 minutes for 25 million tracers, with 44 seconds taken by resolution creation and binning. Reading and writing the multiple time step files (40 in this example case) is the majority of the computational effort. Fetching Trajectories Our interactive flow exploration environment relies heavily on the files created during preprocessing. The coarsest resolution provides the trajectory sub-sampling used for fully interactive pathline interpolation; it is loaded upon initialization. Once a suitable location for the seeding curve or rake is found, refinement begins by loading additional tracers in the vicinity of the rake. These additional tracers are located in finer resolution levels, making it necessary to determine which grid cells are intersected. The reference file provided holds the dimensions and cell sizes for all resolutions. By combining this information with the 3D positions of rake seed points, the cells enclosing the rake can be found and sorted into a list for each resolution. Additionally, the adjacent neighbors of these cells are also sorted into the list by cell index. Using the beginning and end index of the cell list, the memory mapper maps only just enough of the resolution to cover the desired cells, exemplified in Figure 6. While a chunk of memory is mapped, only the desired cells are loaded since the reference file provides exact byte offsets for the file pointer to follow. As new trajectories are added to the pool of known tracers, pathlines are re-interpolated and a new visualization is created. The non-iterative nature of using interpolation for pathreconstruction is readily configured for parallelism (multithreading) during this process. Loading and interpolating position information for every rake seed point is done in parallel. To begin, all known tracers from the coarsest resolution are sorted into a spatial data structure and each seed point locates its nearest neighbors, building its neighborhood. Instead of restructuring the data structure as new tracers are loaded, tracers are added directly to the corresponding neighborhoods and each neighborhood is filtered by a k-nearest neighbor approach. By the end of the refinement process, the rake seed points are surrounded by the k-nearest trajectories over the entire input set of tracers, providing the best possible accuracy for scattered data interpolation Pathline Interpolation For complete flow field exploration, pathlines from arbitrarily seeded particles must be extracted throughout the domain. Such pathlines are interpolated from the input set of tracers provided by the Lagrangian representation. To begin interpolation, the seed location x is found within the current set of tracers at time τ, which form its neighborhood. A weighing kernel is applied to express the seed point as a function of its neighbors, x = f (x j (τ)), where x j are trajectories in the neighborhood. We use the following spline weight function [BMXK06], W(ρ)=1 6ρ 2 + 8ρ 3 3ρ 4 (5) where ρ is the Euclidean distance between x and x j divided by the neighborhood radius. The same function is then applied at a later time step, conditioned by the neighborhood at τ, to interpolate the seed point s new location, x(τ + i)= f (x j (τ + i)). Therefore, given a particle position and its neighborhood of tracers, scattered data interpolation can interpolate the particle s position at any time step during which the tracers exist. For the results presented in this paper, we have chosen to use the Moving Least Squares (MLS) [Lev98] method because of its continuity and local interpolation ability. Figure 5: Reference file. Ordered by resolutions, one integer is written out per cell to indicate how many bytes into the time step files must be skipped to reach the cell s contents. Figure 6: Memory mapped section of file. The rake, shown as a black line, is located within the blue cells. While information is only read from these cells, the entire area in red must be mapped for the file pointer to jump cell to cell, accessing the needed trajectory information.

6 The advantage of interpolation over integration in regards to pathline construction is in the speed of position calculation. For the Eulerian framework, advecting a pathline to a time τ + i involves computationally expensive numerical integration for each previous time step. The position of the particle at times τ +[1,2,...,i 2,i 1] must be calculated before its position at τ +i can be computed. In contrast, only one scattered data reconstruction using the Lagrangian representation is needed to achieve the same position information, thereby lowering the computational cost. We can further reduce the computational cost of pathline representation by choosing the frequency with which position information is calculated. That is, if the goal is to draw a pathline between times τ and τ +i, it may not be necessary to connect the positions of all intermediate time steps to get a sense of the fluid flow in the region, but only a subset of positions. A value κ defines the time interval at which position values are calculated. Therefore, the number of interpolations needed to visually represent a pathline decreases from i to κ i per seeded particle Interactive Data Exploration We demonstrate the advantage of fast position calculation on the CPU by offering an interactive seeding environment to explore the data set. The environment allows flow behavior at any data set position to be viewed in real time, and enables notable flow features such as salient structures to be sought out. To reduce the memory footprint of the exploration environment and to promote real time rake manipulation, only tracers from the coarse resolution are loaded by the algorithm. Because only a subset of tracers are initially loaded, additional trajectories within the rake neighborhood exist in the finer resolutions. Once an area of interest is located, these trajectories are loaded in the vicinity of the rake to improve the interpolation quality of the pathline. In other words, the pathlines are interpolated from the closest possible trajectories to the seed point, providing the best possible accuracy of the reconstruction. The dynamic loading of tracers significantly reduces the amount of I/O when compared to the Eulerian approach. 6. Results and Analysis Let C be a univariate space curve parametrized in s R. A path-surface is the union of all pathlines passing through such a user-defined seeding curve, the rake, where s designates a specific pathline seeded at C(s). Creating visual connections between these curves along the s and t directions creates the path-surface [GKT 08]. If the rake forms a surface instead of a curve, it becomes a two-dimensional family of pathlines with a seed surface S(u,v) R 2, otherwise referred to as a time-surface. Points seeded in this fashion jointly traverse the flow, thus presenting a snapshot of the surface s deformation at a time t i. Stitching these coordinates along the (u, v) axis forms a visual surface through time [KGJ09]. Both types of surfaces can be computed using the trajectory interpolation presented in this paper. For the presented results, the interpolation neighborhood was set to the 12 nearest neighbors and curves created with a value of κ = 4. All computations were done on an Intel Core i7 processor with 7200 RPM and 8GB of RAM. The colors chosen are for illustration purposes only and serve to simply differentiate between interpolated and integrated surfaces Karman Vortex Street We begin by exploring a data set exhibiting a Karman Vortex Street. The given trajectories have a total time step length of 36 and the data set is scaled to fit into a unit cube and 1 contains 1,656 trajectories per 100 of a unit cube, split into 6 levels of resolution. The time step τ chosen for the preprocessing step is randomly selected from the middle of the simulation, giving a snapshot of the flow after a significant amount of time. The Karman data set has one inflow boundary on the left hand side and one outflow boundary on the right where the first quarter of the data set, from left to right, holds a cylindrical object lodged in the flow. As the fluid flows over the cylinder, it pushes downward towards the middle which in turn pushes particles on the side outward and up. Figure 7 exemplifies this trend through a time surface representation. The rake begins as a rectangle, densely packed with points, which is then interpolated through time and visualized at intervals of κ. Here, the rake is positioned above the obstacle showing time surfaces heading downwards. For a close look directly behind the cylinder, an area exhibiting extremely high turbulence, the top of Figure 9 shows three different images of the moving plane in the same location. The first two are reconstructed at the coarse and refined resolutions using our approach, while the final surface is an Eulerian-based advection using adaptive Runge-Kutta 4/5. The first image shows the path-surface reconstruction during interactive exploration, revealing its general shape. The center image shows the surface reconstructed after additional trajectories in the vicinity of the rake are loaded from Figure 7: Time surfaces behind the cylinder. Surfaces are extracted at an interval of κ = 4, showing the downward trend of trajectories from the top towards the bottom.

7 top of the jet. Figure 8 is a path-surface extracted near this plume. The rake is seeded as a circle that captures trajectories spreading out as they curl in on themselves. Figure 8: Path surfaces reconstructed using interpolation (left and middle) and integration (right). The left surface represents a real time interpolated reconstruction from trajectory information, which captures the general shape of the actual surface. The middle image shows how the interpolated reconstruction converges to the same surface when the flow map resolution is increased by dynamic loading of new sample trajectories. The same surface extracted with numerical integration is shown on the right. The color is for illustration purposes only and serves to differentiate the interpolated and integrated surfaces. all levels of resolution. The increased density of trajectories drastically improves the accuracy of the reconstruction. Below each image in Figure 9 are reconstruction metrics measuring accuracy, performance, and I/O. Accuracy is measured as the Euclidean distance between the trajectories that make up the surface reconstructed through interpolation and the surface created through numerical integration. We treat the advected surface as the ground truth and show that our method can achieve comparable accuracy. The left-most and center image show that as additional trajectories are loaded, the accuracy improves and the difference in surfaces, when compared to the ground truth, diminishes. While not as accurate, the coarse reconstruction allows for a surface preview with the benefit of real-time visualization. To match ground truth accuracy, loading additional tracers takes 1,143ms, which is slightly above half the time it takes for the Eulerian surface (1,936ms). Additional trajectories are only loaded in the vicinity of the plane, resulting in a total of 25.7MB necessary to perform the reconstruction. When compared to the 84.8MB of velocity field information needed to perform advection, our approach uses over 3 times less memory to create a near identical surface. Moreover, our method also takes less time to construct the surface, reaching comparable accuracy faster and with less memory than the Eulerian representation The Jet We continue with a look at a "Jet" data set. The given trajectories have a total time step length of 30 and the data set is scaled to fit into a unit cube and contains 1,656 trajectories 1 per 100 of a unit cube, split into 4 levels of resolution. An overall turbulent data set, the Jet has an inflow boundary at the bottom in the form of four jets at the center. A prominent feature of interest is the plume that forms towards the To explore the plume in further detail, we switch to a timesurface to better capture the curl of the structure. The bottom of Figure 9 shows three surfaces, two of which are extracted using interpolation and the final using numerical integration. All images look up at the underside of the forming plume, capturing it folding in on itself. As additional trajectories are loaded from finer resolutions, the interpolated surface matches up visually to the integrated surface. This is shown numerically through the average distance error metric below the images, which decreases with refinement. For a time surface, refinement loading is constant because position information is required at only one time step. With numerical integration however, all seed points must be advected to the representative time, thereby making computation time linear. The time-surface was extracted after 30 time steps, taking numerical integration 10,941ms, while completely refined interpolation took only 2,362ms. The results also shows that through the refinement step, the Lagrangian representation loaded 27.4MB of memory, compared to the 737MB loaded with numerical integration. Our method only needs to load 2 time steps when extracting time surfaces, which significantly reduces the computation time and I/O of creating the surface when compared with advection with the Eulerian representation. 7. Concluding Remarks The main contribution of this work is a novel file system together with interpolation techniques that provide significant advantages over classic time-varying integration, such as direct control over memory constraints. We show that tracer interpolation reduces computational effort in such a way as to provide a real-time, interactive tool for exploring fluid flow. While the overall process involves a pre-processing step for a given time step τ, this step only needs to be performed once to gain the benefits of a Lagrangian flow map representation over the entire data set. For future work, we plan to extend the pre-processing stage to include multiple seeding time steps and support streak surface interpolation. References [BFTW09] BURGER K., FERSTL F., THEISEL H., WESTER- MANN R.: Interactive streak surface visualization on the GPU. IEEE Transactions on Visualization and Computer Graphics 15, 6 (2009), [BMXK06] BODIN A., MA J., XIN X., KRISHNASWAMI P.: A meshless integral method based on regularized boundary integral equation. Computer Methods in Applied Mechanics and Engineering, (2006). 5 [CGC 11] CAMP D., GARTH C., CHILDS H., PUGMIRE D., JOY K.: Streamline integration using mpi-hybrid parallelism on a large multicore architecture. IEEE Transactions on Visualization and Computer Graphics 17, 11 (2011), , 2

8 Figure 9: Path-surface behind the cylinder (Karman) and time-surface of the plume (Jet) extracted with 100 seeds over 36 steps and 10,000 seeds over 30 steps respectively. Surfaces reconstructed with interpolation (left and middle) and integration (right). [CK90] CASH J. R., KARP A. H.: A variable order runge-kutta method for initial value problems with rapidly varying right-hand sides. ACM Trans. Math. Softw. 16 (1990), [GGTH07] GARTH C., GERHARDT F., TRICOCHE X., HAGEN H.: Efficient computation and visualization of coherent structures in fluid flow applications. IEEE Transactions on Visualization and Computer Graphics 13, 6 (2007), [GKT 08] GARTH C., KRISHNAN H., TRICOCHE X., BOBACH T., JOY K.: Generation of accurate integral surfaces in timedependent vector fields. IEEE Transactions on Visualization and Computer Graphics 14, 6 (2008), , 6 [Hal01] HALLER G.: Distinguished material surfaces and coherent structures in three-dimensional fluid flows. Physica D: Nonlinear Phenomena 149, 4 (2001), , 2 [HSW11] HLAWATSCH M., SADLO F., WEISKOPF D.: Hierarchical line integration. IEEE Transactions on Visualization and Computer Graphics 17, 8 (2011), [Hul92] HULTQUIST J. P. M.: Constructing stream surfaces in steady 3d vector fields. In Proceedings of IEEE Visualization 1992 (1992), pp [KGJ09] KRISHNAN H., GARTH C., JOY K.: Time and streak surfaces for flow visualization in large time-varying data sets. IEEE Transactions on Visualization and Computer Graphics 15, 6 (2009), , 6 [Lev98] LEVIN D.: The approximation power of moving leastsquares. Mathematics of Computation 67 (1998), [LRR00] LODHA S., RENTERIA J., ROSKIN K.: Topology preserving compression of 2d vector fields. In Proceedings of Visualization 2000 (2000), pp [MLP 10] MCLOUGLIN T., LARAMEE R. S., PEIKERT R., POST F. H., CHEN M.: Over two decades of integration-based geometric flow visualization. Computer Graphics Forum 29, 6 (2010), [Mon88] MONAGHAN J.: An introduction to SPH. Computer Physics Communications 48, 1 (1988), [OHBH13] OBERMAIER H., HERING-BERTRAM M., HAGEN H.: Topological flow volume extraction from time-surface maps. Computer Aided Geometric Design 30, 6 (2013), [SP07] SADLO F., PEIKERT R.: Efficient visualization of lagrangian coherent structures by filtered AMR ridge extraction. IEEE Transactions on Visualization and Computer Graphics 13, 6 (2007), [SWS07] SALZBRUNN T., WIEBEL A., SCHEUERMANN G.: Efficient construction of flow structures. In The Seventh IASTED International Conference on Visualization, Imaging and Image Processing (2007), pp [TS03] THEISEL H., SEIDEL H.-P.: Feature flow fields. In Proceedings of the Symposium on Data Visualisation (2003), pp [vw93] VAN WIJK J. J.: Implicit stream surfaces. In VIS 93: Proceedings of the Conference on Visualization 93 (1993), pp [WJE00] WESTERMANN R., JOHNSON C., ERTL T.: A level-set method for flow visualization. In Proceedings of IEEE Visualization 2000 (2000), pp [WT10] WEINKAUF T., THEISEL H.: Streak lines as tangent curves of a derived vector field. IEEE Transactions on Visualization and Computer Graphics 16, 6 (2010),

A Multi-Resolution Interpolation Scheme for Pathline Based Lagrangian Flow Representations

A Multi-Resolution Interpolation Scheme for Pathline Based Lagrangian Flow Representations A Multi-Resolution Interpolation Scheme for Pathline Based Lagrangian Flow Representations Alexy Agranovsky a, Harald Obermaier a, Christoph Garth b, and Kenneth I. Joy a a University of California, Davis,

More information

Part I: Theoretical Background and Integration-Based Methods

Part I: Theoretical Background and Integration-Based Methods Large Vector Field Visualization: Theory and Practice Part I: Theoretical Background and Integration-Based Methods Christoph Garth Overview Foundations Time-Varying Vector Fields Numerical Integration

More information

Flow Visualization with Integral Surfaces

Flow Visualization with Integral Surfaces Flow Visualization with Integral Surfaces Visual and Interactive Computing Group Department of Computer Science Swansea University R.S.Laramee@swansea.ac.uk 1 1 Overview Flow Visualization with Integral

More information

Using Integral Surfaces to Visualize CFD Data

Using Integral Surfaces to Visualize CFD Data Using Integral Surfaces to Visualize CFD Data Tony Mcloughlin, Matthew Edmunds,, Mark W. Jones, Guoning Chen, Eugene Zhang 1 1 Overview Flow Visualization with Integral Surfaces: Introduction to flow visualization

More information

Implicit Integral Surfaces

Implicit Integral Surfaces Vision, Modeling, and Visualization (2012) M. Goesele, T. Grosch, B. Preim, H. Theisel, and K. Toennies (Eds.) Implicit Integral Surfaces T. Stöter 1 and T. Weinkauf 1 and H.-P. Seidel 1 and H. Theisel

More information

Extracting Flow Structures Using Sparse Particles

Extracting Flow Structures Using Sparse Particles Vision, Modeling, and Visualization (2011) Peter Eisert, Konrad Polthier, and Joachim Hornegger (Eds.) Extracting Flow Structures Using Sparse Particles Alexy Agranovsky, Christoph Garth, and Kenneth I.

More information

Post Processing, Visualization, and Sample Output

Post Processing, Visualization, and Sample Output Chapter 7 Post Processing, Visualization, and Sample Output Upon successful execution of an ADCIRC run, a number of output files will be created. Specifically which files are created depends upon how the

More information

Vector Field Visualisation

Vector Field Visualisation Vector Field Visualisation Computer Animation and Visualization Lecture 14 Institute for Perception, Action & Behaviour School of Informatics Visualising Vectors Examples of vector data: meteorological

More information

Acknowledgements. Prof. Dan Negrut Prof. Darryl Thelen Prof. Michael Zinn. SBEL Colleagues: Hammad Mazar, Toby Heyn, Manoj Kumar

Acknowledgements. Prof. Dan Negrut Prof. Darryl Thelen Prof. Michael Zinn. SBEL Colleagues: Hammad Mazar, Toby Heyn, Manoj Kumar Philipp Hahn Acknowledgements Prof. Dan Negrut Prof. Darryl Thelen Prof. Michael Zinn SBEL Colleagues: Hammad Mazar, Toby Heyn, Manoj Kumar 2 Outline Motivation Lumped Mass Model Model properties Simulation

More information

Overview of Traditional Surface Tracking Methods

Overview of Traditional Surface Tracking Methods Liquid Simulation With Mesh-Based Surface Tracking Overview of Traditional Surface Tracking Methods Matthias Müller Introduction Research lead of NVIDIA PhysX team PhysX GPU acc. Game physics engine www.nvidia.com\physx

More information

EVALUATING SPATIOTEMPORAL SEARCH STRUCTURES FOR ANALYSIS FROM LAGRANGIAN BASIS FLOWS. by: Ouermi Timbwaoaga Aime Judicael (TAJO)

EVALUATING SPATIOTEMPORAL SEARCH STRUCTURES FOR ANALYSIS FROM LAGRANGIAN BASIS FLOWS. by: Ouermi Timbwaoaga Aime Judicael (TAJO) EVALUATING SPATIOTEMPORAL SEARCH STRUCTURES FOR ANALYSIS FROM LAGRANGIAN BASIS FLOWS by: Ouermi Timbwaoaga Aime Judicael (TAJO) Presented to the Department of Computer and Information Science at the University

More information

Computer Graphics Ray Casting. Matthias Teschner

Computer Graphics Ray Casting. Matthias Teschner Computer Graphics Ray Casting Matthias Teschner Outline Context Implicit surfaces Parametric surfaces Combined objects Triangles Axis-aligned boxes Iso-surfaces in grids Summary University of Freiburg

More information

Lagrangian methods and Smoothed Particle Hydrodynamics (SPH) Computation in Astrophysics Seminar (Spring 2006) L. J. Dursi

Lagrangian methods and Smoothed Particle Hydrodynamics (SPH) Computation in Astrophysics Seminar (Spring 2006) L. J. Dursi Lagrangian methods and Smoothed Particle Hydrodynamics (SPH) Eulerian Grid Methods The methods covered so far in this course use an Eulerian grid: Prescribed coordinates In `lab frame' Fluid elements flow

More information

1.2 Numerical Solutions of Flow Problems

1.2 Numerical Solutions of Flow Problems 1.2 Numerical Solutions of Flow Problems DIFFERENTIAL EQUATIONS OF MOTION FOR A SIMPLIFIED FLOW PROBLEM Continuity equation for incompressible flow: 0 Momentum (Navier-Stokes) equations for a Newtonian

More information

Multigrid Pattern. I. Problem. II. Driving Forces. III. Solution

Multigrid Pattern. I. Problem. II. Driving Forces. III. Solution Multigrid Pattern I. Problem Problem domain is decomposed into a set of geometric grids, where each element participates in a local computation followed by data exchanges with adjacent neighbors. The grids

More information

Over Two Decades of IntegrationBased, Geometric Vector Field. Visualization

Over Two Decades of IntegrationBased, Geometric Vector Field. Visualization Over Two Decades of IntegrationBased, Geometric Vector Field Visualization Tony McLoughlin1, 1, Ronald Peikert2, Frits H. Post3, and Min Chen1 1 The Visual and Interactive Computing Group Computer Science

More information

Vector Visualization

Vector Visualization Vector Visualization Vector Visulization Divergence and Vorticity Vector Glyphs Vector Color Coding Displacement Plots Stream Objects Texture-Based Vector Visualization Simplified Representation of Vector

More information

3D vector fields. Contents. Introduction 3D vector field topology Representation of particle lines. 3D LIC Combining different techniques

3D vector fields. Contents. Introduction 3D vector field topology Representation of particle lines. 3D LIC Combining different techniques 3D vector fields Scientific Visualization (Part 9) PD Dr.-Ing. Peter Hastreiter Contents Introduction 3D vector field topology Representation of particle lines Path lines Ribbons Balls Tubes Stream tetrahedra

More information

Lecture overview. Visualisatie BMT. Vector algorithms. Vector algorithms. Time animation. Time animation

Lecture overview. Visualisatie BMT. Vector algorithms. Vector algorithms. Time animation. Time animation Visualisatie BMT Lecture overview Vector algorithms Tensor algorithms Modeling algorithms Algorithms - 2 Arjan Kok a.j.f.kok@tue.nl 1 2 Vector algorithms Vector 2 or 3 dimensional representation of direction

More information

Vector Visualisation 1. global view

Vector Visualisation 1. global view Vector Field Visualisation : global view Visualisation Lecture 12 Institute for Perception, Action & Behaviour School of Informatics Vector Visualisation 1 Vector Field Visualisation : local & global Vector

More information

GPU Acceleration of Particle Advection Workloads in a Parallel, Distributed Memory Setting

GPU Acceleration of Particle Advection Workloads in a Parallel, Distributed Memory Setting Girona, Spain May 4-5 GPU Acceleration of Particle Advection Workloads in a Parallel, Distributed Memory Setting David Camp, Hari Krishnan, David Pugmire, Christoph Garth, Ian Johnson, E. Wes Bethel, Kenneth

More information

An Introduction to Flow Visualization (1) Christoph Garth

An Introduction to Flow Visualization (1) Christoph Garth An Introduction to Flow Visualization (1) Christoph Garth cgarth@ucdavis.edu Motivation What will I be talking about? Classical: Physical experiments to understand flow. 2 Motivation What will I be talking

More information

Texture Advection. Ronald Peikert SciVis Texture Advection 6-1

Texture Advection. Ronald Peikert SciVis Texture Advection 6-1 Texture Advection Ronald Peikert SciVis 2007 - Texture Advection 6-1 Texture advection Motivation: dense visualization of vector fields, no seed points needed. Methods for static fields: LIC - Line integral

More information

Data Visualization. Fall 2017

Data Visualization. Fall 2017 Data Visualization Fall 2017 Vector Fields Vector field v: D R n D is typically 2D planar surface or 2D surface embedded in 3D n = 2 fields tangent to 2D surface n = 3 volumetric fields When visualizing

More information

A Toolbox of Level Set Methods

A Toolbox of Level Set Methods A Toolbox of Level Set Methods Ian Mitchell Department of Computer Science University of British Columbia http://www.cs.ubc.ca/~mitchell mitchell@cs.ubc.ca research supported by the Natural Science and

More information

ALE Seamless Immersed Boundary Method with Overset Grid System for Multiple Moving Objects

ALE Seamless Immersed Boundary Method with Overset Grid System for Multiple Moving Objects Tenth International Conference on Computational Fluid Dynamics (ICCFD10), Barcelona,Spain, July 9-13, 2018 ICCFD10-047 ALE Seamless Immersed Boundary Method with Overset Grid System for Multiple Moving

More information

Lagrangian and Eulerian Representations of Fluid Flow: Kinematics and the Equations of Motion

Lagrangian and Eulerian Representations of Fluid Flow: Kinematics and the Equations of Motion Lagrangian and Eulerian Representations of Fluid Flow: Kinematics and the Equations of Motion James F. Price Woods Hole Oceanographic Institution Woods Hole, MA, 02543 July 31, 2006 Summary: This essay

More information

Supplemental Material Deep Fluids: A Generative Network for Parameterized Fluid Simulations

Supplemental Material Deep Fluids: A Generative Network for Parameterized Fluid Simulations Supplemental Material Deep Fluids: A Generative Network for Parameterized Fluid Simulations 1. Extended Results 1.1. 2-D Smoke Plume Additional results for the 2-D smoke plume example are shown in Figures

More information

CGT 581 G Fluids. Overview. Some terms. Some terms

CGT 581 G Fluids. Overview. Some terms. Some terms CGT 581 G Fluids Bedřich Beneš, Ph.D. Purdue University Department of Computer Graphics Technology Overview Some terms Incompressible Navier-Stokes Boundary conditions Lagrange vs. Euler Eulerian approaches

More information

CFD MODELING FOR PNEUMATIC CONVEYING

CFD MODELING FOR PNEUMATIC CONVEYING CFD MODELING FOR PNEUMATIC CONVEYING Arvind Kumar 1, D.R. Kaushal 2, Navneet Kumar 3 1 Associate Professor YMCAUST, Faridabad 2 Associate Professor, IIT, Delhi 3 Research Scholar IIT, Delhi e-mail: arvindeem@yahoo.co.in

More information

Driven Cavity Example

Driven Cavity Example BMAppendixI.qxd 11/14/12 6:55 PM Page I-1 I CFD Driven Cavity Example I.1 Problem One of the classic benchmarks in CFD is the driven cavity problem. Consider steady, incompressible, viscous flow in a square

More information

DRP Report: Accelerating Advection Via Approximate Block Exterior Flow Maps

DRP Report: Accelerating Advection Via Approximate Block Exterior Flow Maps DRP Report: Accelerating Advection Via Approximate Block Exterior Flow Maps Ryan Bleile University of Oregon ABSTRACT Flow visualization techniques involving extreme advection workloads are becoming increasingly

More information

Ian Mitchell. Department of Computer Science The University of British Columbia

Ian Mitchell. Department of Computer Science The University of British Columbia CPSC 542D: Level Set Methods Dynamic Implicit Surfaces and the Hamilton-Jacobi Equation or What Water Simulation, Robot Path Planning and Aircraft Collision Avoidance Have in Common Ian Mitchell Department

More information

Skåne University Hospital Lund, Lund, Sweden 2 Deparment of Numerical Analysis, Centre for Mathematical Sciences, Lund University, Lund, Sweden

Skåne University Hospital Lund, Lund, Sweden 2 Deparment of Numerical Analysis, Centre for Mathematical Sciences, Lund University, Lund, Sweden Volume Tracking: A New Method for Visualization of Intracardiac Blood Flow from Three-Dimensional, Time-Resolved, Three-Component Magnetic Resonance Velocity Mapping Appendix: Theory and Numerical Implementation

More information

Simulation of liquid cube fracture with SPH

Simulation of liquid cube fracture with SPH Journal of Physics: Conference Series PAPER OPEN ACCESS Simulation of liquid cube fracture with SPH To cite this article: M N Davydov 2016 J. Phys.: Conf. Ser. 754 062001 View the article online for updates

More information

Data Partitioning. Figure 1-31: Communication Topologies. Regular Partitions

Data Partitioning. Figure 1-31: Communication Topologies. Regular Partitions Data In single-program multiple-data (SPMD) parallel programs, global data is partitioned, with a portion of the data assigned to each processing node. Issues relevant to choosing a partitioning strategy

More information

Vector Visualization. CSC 7443: Scientific Information Visualization

Vector Visualization. CSC 7443: Scientific Information Visualization Vector Visualization Vector data A vector is an object with direction and length v = (v x,v y,v z ) A vector field is a field which associates a vector with each point in space The vector data is 3D representation

More information

Realtime Water Simulation on GPU. Nuttapong Chentanez NVIDIA Research

Realtime Water Simulation on GPU. Nuttapong Chentanez NVIDIA Research 1 Realtime Water Simulation on GPU Nuttapong Chentanez NVIDIA Research 2 3 Overview Approaches to realtime water simulation Hybrid shallow water solver + particles Hybrid 3D tall cell water solver + particles

More information

Vector Field Visualization: Introduction

Vector Field Visualization: Introduction Vector Field Visualization: Introduction What is a Vector Field? A simple 2D steady vector field A vector valued function that assigns a vector (with direction and magnitude) to any given point. It typically

More information

Level set methods Formulation of Interface Propagation Boundary Value PDE Initial Value PDE Motion in an externally generated velocity field

Level set methods Formulation of Interface Propagation Boundary Value PDE Initial Value PDE Motion in an externally generated velocity field Level Set Methods Overview Level set methods Formulation of Interface Propagation Boundary Value PDE Initial Value PDE Motion in an externally generated velocity field Convection Upwind ddifferencingi

More information

Volume Illumination & Vector Field Visualisation

Volume Illumination & Vector Field Visualisation Volume Illumination & Vector Field Visualisation Visualisation Lecture 11 Institute for Perception, Action & Behaviour School of Informatics Volume Illumination & Vector Vis. 1 Previously : Volume Rendering

More information

Module 1 Lecture Notes 2. Optimization Problem and Model Formulation

Module 1 Lecture Notes 2. Optimization Problem and Model Formulation Optimization Methods: Introduction and Basic concepts 1 Module 1 Lecture Notes 2 Optimization Problem and Model Formulation Introduction In the previous lecture we studied the evolution of optimization

More information

03 - Reconstruction. Acknowledgements: Olga Sorkine-Hornung. CSCI-GA Geometric Modeling - Spring 17 - Daniele Panozzo

03 - Reconstruction. Acknowledgements: Olga Sorkine-Hornung. CSCI-GA Geometric Modeling - Spring 17 - Daniele Panozzo 3 - Reconstruction Acknowledgements: Olga Sorkine-Hornung Geometry Acquisition Pipeline Scanning: results in range images Registration: bring all range images to one coordinate system Stitching/ reconstruction:

More information

CS123 INTRODUCTION TO COMPUTER GRAPHICS. Describing Shapes. Constructing Objects in Computer Graphics 1/15

CS123 INTRODUCTION TO COMPUTER GRAPHICS. Describing Shapes. Constructing Objects in Computer Graphics 1/15 Describing Shapes Constructing Objects in Computer Graphics 1/15 2D Object Definition (1/3) Lines and polylines: Polylines: lines drawn between ordered points A closed polyline is a polygon, a simple polygon

More information

Numerical Simulation of Coastal Wave Processes with the Use of Smoothed Particle Hydrodynamics (SPH) Method

Numerical Simulation of Coastal Wave Processes with the Use of Smoothed Particle Hydrodynamics (SPH) Method Aristotle University of Thessaloniki Faculty of Engineering Department of Civil Engineering Division of Hydraulics and Environmental Engineering Laboratory of Maritime Engineering Christos V. Makris Dipl.

More information

CFD Project Workflow Guide

CFD Project Workflow Guide CFD Project Workflow Guide Contents Select a problem with known results for proof-of-concept testing... 1 Set up and run a coarse test case... 2 Select and calibrate numerical methods... 3 Minimize & quantify

More information

Lecture 17: Solid Modeling.... a cubit on the one side, and a cubit on the other side Exodus 26:13

Lecture 17: Solid Modeling.... a cubit on the one side, and a cubit on the other side Exodus 26:13 Lecture 17: Solid Modeling... a cubit on the one side, and a cubit on the other side Exodus 26:13 Who is on the LORD's side? Exodus 32:26 1. Solid Representations A solid is a 3-dimensional shape with

More information

Intersection Acceleration

Intersection Acceleration Advanced Computer Graphics Intersection Acceleration Matthias Teschner Computer Science Department University of Freiburg Outline introduction bounding volume hierarchies uniform grids kd-trees octrees

More information

weighted minimal surface model for surface reconstruction from scattered points, curves, and/or pieces of surfaces.

weighted minimal surface model for surface reconstruction from scattered points, curves, and/or pieces of surfaces. weighted minimal surface model for surface reconstruction from scattered points, curves, and/or pieces of surfaces. joint work with (S. Osher, R. Fedkiw and M. Kang) Desired properties for surface reconstruction:

More information

WAVE PATTERNS, WAVE INDUCED FORCES AND MOMENTS FOR A GRAVITY BASED STRUCTURE PREDICTED USING CFD

WAVE PATTERNS, WAVE INDUCED FORCES AND MOMENTS FOR A GRAVITY BASED STRUCTURE PREDICTED USING CFD Proceedings of the ASME 2011 30th International Conference on Ocean, Offshore and Arctic Engineering OMAE2011 June 19-24, 2011, Rotterdam, The Netherlands OMAE2011-49593 WAVE PATTERNS, WAVE INDUCED FORCES

More information

Ray Tracing Acceleration Data Structures

Ray Tracing Acceleration Data Structures Ray Tracing Acceleration Data Structures Sumair Ahmed October 29, 2009 Ray Tracing is very time-consuming because of the ray-object intersection calculations. With the brute force method, each ray has

More information

Vector Field Visualization: Introduction

Vector Field Visualization: Introduction Vector Field Visualization: Introduction What is a Vector Field? Why It is Important? Vector Fields in Engineering and Science Automotive design [Chen et al. TVCG07,TVCG08] Weather study [Bhatia and Chen

More information

An Integral Curve Attribute Based Flow Segmentation

An Integral Curve Attribute Based Flow Segmentation Noname manuscript No. (will be inserted by the editor) An Integral Curve Attribute Based Flow Segmentation Lei Zhang, Robert S. Laramee, David Thompson, Adrian Sescu, and Guoning Chen Received: date /

More information

Non-Homogeneous Swarms vs. MDP s A Comparison of Path Finding Under Uncertainty

Non-Homogeneous Swarms vs. MDP s A Comparison of Path Finding Under Uncertainty Non-Homogeneous Swarms vs. MDP s A Comparison of Path Finding Under Uncertainty Michael Comstock December 6, 2012 1 Introduction This paper presents a comparison of two different machine learning systems

More information

An explicit and conservative remapping strategy for semi-lagrangian advection

An explicit and conservative remapping strategy for semi-lagrangian advection An explicit and conservative remapping strategy for semi-lagrangian advection Sebastian Reich Universität Potsdam, Potsdam, Germany January 17, 2007 Abstract A conservative semi-lagrangian advection scheme

More information

Adaptive-Mesh-Refinement Pattern

Adaptive-Mesh-Refinement Pattern Adaptive-Mesh-Refinement Pattern I. Problem Data-parallelism is exposed on a geometric mesh structure (either irregular or regular), where each point iteratively communicates with nearby neighboring points

More information

An Eulerian Approach for Computing the Finite Time Lyapunov Exponent (FTLE)

An Eulerian Approach for Computing the Finite Time Lyapunov Exponent (FTLE) An Eulerian Approach for Computing the Finite Time Lyapunov Exponent (FTLE) Shingyu Leung Department of Mathematics, Hong Kong University of Science and Technology masyleung@ust.hk May, Shingyu Leung (HKUST)

More information

CS 563 Advanced Topics in Computer Graphics QSplat. by Matt Maziarz

CS 563 Advanced Topics in Computer Graphics QSplat. by Matt Maziarz CS 563 Advanced Topics in Computer Graphics QSplat by Matt Maziarz Outline Previous work in area Background Overview In-depth look File structure Performance Future Point Rendering To save on setup and

More information

Vector Field Visualization: Computing Topology

Vector Field Visualization: Computing Topology Vector Field Visualization: Computing Topology Vector Field Topology (Recall) Vector field topology provides qualitative (structural) information of the underlying dynamics It usually consists of certain

More information

Discrete representations of geometric objects: Features, data structures and adequacy for dynamic simulation. Part I : Solid geometry

Discrete representations of geometric objects: Features, data structures and adequacy for dynamic simulation. Part I : Solid geometry Discrete representations of geometric objects: Features, data structures and adequacy for dynamic simulation. Surfaces Part I : Solid geometry hachar Fleishman Tel Aviv University David Levin Claudio T.

More information

Contextual Geometric Structures

Contextual Geometric Structures Contextual Geometric Structures Modeling the fundamental components of cultural behavior Bradly Alicea http://www.msu.edu/~aliceabr/ What is the essence of culture? Heredity (beliefs that propagate)? Plasticity?

More information

Aerodynamic Study of a Realistic Car W. TOUGERON

Aerodynamic Study of a Realistic Car W. TOUGERON Aerodynamic Study of a Realistic Car W. TOUGERON Tougeron CFD Engineer 2016 Abstract This document presents an aerodynamic CFD study of a realistic car geometry. The aim is to demonstrate the efficiency

More information

Many rendering scenarios, such as battle scenes or urban environments, require rendering of large numbers of autonomous characters.

Many rendering scenarios, such as battle scenes or urban environments, require rendering of large numbers of autonomous characters. 1 2 Many rendering scenarios, such as battle scenes or urban environments, require rendering of large numbers of autonomous characters. Crowd rendering in large environments presents a number of challenges,

More information

Realistic Animation of Fluids

Realistic Animation of Fluids 1 Realistic Animation of Fluids Nick Foster and Dimitris Metaxas Presented by Alex Liberman April 19, 2005 2 Previous Work Used non physics-based methods (mostly in 2D) Hard to simulate effects that rely

More information

Space Filling Curves and Hierarchical Basis. Klaus Speer

Space Filling Curves and Hierarchical Basis. Klaus Speer Space Filling Curves and Hierarchical Basis Klaus Speer Abstract Real world phenomena can be best described using differential equations. After linearisation we have to deal with huge linear systems of

More information

G 2 Interpolation for Polar Surfaces

G 2 Interpolation for Polar Surfaces 1 G 2 Interpolation for Polar Surfaces Jianzhong Wang 1, Fuhua Cheng 2,3 1 University of Kentucky, jwangf@uky.edu 2 University of Kentucky, cheng@cs.uky.edu 3 National Tsinhua University ABSTRACT In this

More information

Fairing Scalar Fields by Variational Modeling of Contours

Fairing Scalar Fields by Variational Modeling of Contours Fairing Scalar Fields by Variational Modeling of Contours Martin Bertram University of Kaiserslautern, Germany Abstract Volume rendering and isosurface extraction from three-dimensional scalar fields are

More information

Using GPUs to compute the multilevel summation of electrostatic forces

Using GPUs to compute the multilevel summation of electrostatic forces Using GPUs to compute the multilevel summation of electrostatic forces David J. Hardy Theoretical and Computational Biophysics Group Beckman Institute for Advanced Science and Technology University of

More information

9. Three Dimensional Object Representations

9. Three Dimensional Object Representations 9. Three Dimensional Object Representations Methods: Polygon and Quadric surfaces: For simple Euclidean objects Spline surfaces and construction: For curved surfaces Procedural methods: Eg. Fractals, Particle

More information

iric Software Changing River Science River2D Tutorials

iric Software Changing River Science River2D Tutorials iric Software Changing River Science River2D Tutorials iric Software Changing River Science Confluence of the Colorado River, Blue River and Indian Creek, Colorado, USA 1 TUTORIAL 1: RIVER2D STEADY SOLUTION

More information

A nodal based evolutionary structural optimisation algorithm

A nodal based evolutionary structural optimisation algorithm Computer Aided Optimum Design in Engineering IX 55 A dal based evolutionary structural optimisation algorithm Y.-M. Chen 1, A. J. Keane 2 & C. Hsiao 1 1 ational Space Program Office (SPO), Taiwan 2 Computational

More information

A Singular Example for the Averaged Mean Curvature Flow

A Singular Example for the Averaged Mean Curvature Flow To appear in Experimental Mathematics Preprint Vol. No. () pp. 3 7 February 9, A Singular Example for the Averaged Mean Curvature Flow Uwe F. Mayer Abstract An embedded curve is presented which under numerical

More information

Introduction to Indexing R-trees. Hong Kong University of Science and Technology

Introduction to Indexing R-trees. Hong Kong University of Science and Technology Introduction to Indexing R-trees Dimitris Papadias Hong Kong University of Science and Technology 1 Introduction to Indexing 1. Assume that you work in a government office, and you maintain the records

More information

CS 465 Program 4: Modeller

CS 465 Program 4: Modeller CS 465 Program 4: Modeller out: 30 October 2004 due: 16 November 2004 1 Introduction In this assignment you will work on a simple 3D modelling system that uses simple primitives and curved surfaces organized

More information

An Efficient, Geometric Multigrid Solver for the Anisotropic Diffusion Equation in Two and Three Dimensions

An Efficient, Geometric Multigrid Solver for the Anisotropic Diffusion Equation in Two and Three Dimensions 1 n Efficient, Geometric Multigrid Solver for the nisotropic Diffusion Equation in Two and Three Dimensions Tolga Tasdizen, Ross Whitaker UUSCI-2004-002 Scientific Computing and Imaging Institute University

More information

Flow Visualisation - Background. CITS4241 Visualisation Lectures 20 and 21

Flow Visualisation - Background. CITS4241 Visualisation Lectures 20 and 21 CITS4241 Visualisation Lectures 20 and 21 Flow Visualisation Flow visualisation is important in both science and engineering From a "theoretical" study of o turbulence or o a fusion reactor plasma, to

More information

Particle-based Fluid Simulation

Particle-based Fluid Simulation Simulation in Computer Graphics Particle-based Fluid Simulation Matthias Teschner Computer Science Department University of Freiburg Application (with Pixar) 10 million fluid + 4 million rigid particles,

More information

University of Florida CISE department Gator Engineering. Clustering Part 4

University of Florida CISE department Gator Engineering. Clustering Part 4 Clustering Part 4 Dr. Sanjay Ranka Professor Computer and Information Science and Engineering University of Florida, Gainesville DBSCAN DBSCAN is a density based clustering algorithm Density = number of

More information

Visualization Computer Graphics I Lecture 20

Visualization Computer Graphics I Lecture 20 15-462 Computer Graphics I Lecture 20 Visualization Height Fields and Contours Scalar Fields Volume Rendering Vector Fields [Angel Ch. 12] November 20, 2003 Doug James Carnegie Mellon University http://www.cs.cmu.edu/~djames/15-462/fall03

More information

Example 13 - Shock Tube

Example 13 - Shock Tube Example 13 - Shock Tube Summary This famous experiment is interesting for observing the shock-wave propagation. Moreover, this case uses the representation of perfect gas and compares the different formulations:

More information

Rotating Moving Boundary Analysis Using ANSYS 5.7

Rotating Moving Boundary Analysis Using ANSYS 5.7 Abstract Rotating Moving Boundary Analysis Using ANSYS 5.7 Qin Yin Fan CYBERNET SYSTEMS CO., LTD. Rich Lange ANSYS Inc. As subroutines in commercial software, APDL (ANSYS Parametric Design Language) provides

More information

What is Multigrid? They have been extended to solve a wide variety of other problems, linear and nonlinear.

What is Multigrid? They have been extended to solve a wide variety of other problems, linear and nonlinear. AMSC 600/CMSC 760 Fall 2007 Solution of Sparse Linear Systems Multigrid, Part 1 Dianne P. O Leary c 2006, 2007 What is Multigrid? Originally, multigrid algorithms were proposed as an iterative method to

More information

Chapter 1 - Basic Equations

Chapter 1 - Basic Equations 2.20 Marine Hydrodynamics, Fall 2017 Lecture 2 Copyright c 2017 MIT - Department of Mechanical Engineering, All rights reserved. 2.20 Marine Hydrodynamics Lecture 2 Chapter 1 - Basic Equations 1.1 Description

More information

1 Mathematical Concepts

1 Mathematical Concepts 1 Mathematical Concepts Mathematics is the language of geophysical fluid dynamics. Thus, in order to interpret and communicate the motions of the atmosphere and oceans. While a thorough discussion of the

More information

Development of a Maxwell Equation Solver for Application to Two Fluid Plasma Models. C. Aberle, A. Hakim, and U. Shumlak

Development of a Maxwell Equation Solver for Application to Two Fluid Plasma Models. C. Aberle, A. Hakim, and U. Shumlak Development of a Maxwell Equation Solver for Application to Two Fluid Plasma Models C. Aberle, A. Hakim, and U. Shumlak Aerospace and Astronautics University of Washington, Seattle American Physical Society

More information

Applications of ICFD /SPH Solvers by LS-DYNA to Solve Water Splashing Impact to Automobile Body. Abstract

Applications of ICFD /SPH Solvers by LS-DYNA to Solve Water Splashing Impact to Automobile Body. Abstract Applications of ICFD /SPH Solvers by LS-DYNA to Solve Water Splashing Impact to Automobile Body George Wang (1 ), Kevin Gardner (3), Eric DeHoff (1), Facundo del Pin (2), Inaki Caldichoury (2), Edouard

More information

Tutorial 6. Pumping Well and River

Tutorial 6. Pumping Well and River Tutorial 6 Pumping Well and River Table of Contents Objective. 1 Step-by-Step Procedure... 2 Section 1 Data Input. 2 Step 1: Open Adaptive Groundwater Input (.agw) File. 2 Step 2: Pumping Well Design Database

More information

Clustering Part 4 DBSCAN

Clustering Part 4 DBSCAN Clustering Part 4 Dr. Sanjay Ranka Professor Computer and Information Science and Engineering University of Florida, Gainesville DBSCAN DBSCAN is a density based clustering algorithm Density = number of

More information

Adarsh Krishnamurthy (cs184-bb) Bela Stepanova (cs184-bs)

Adarsh Krishnamurthy (cs184-bb) Bela Stepanova (cs184-bs) OBJECTIVE FLUID SIMULATIONS Adarsh Krishnamurthy (cs184-bb) Bela Stepanova (cs184-bs) The basic objective of the project is the implementation of the paper Stable Fluids (Jos Stam, SIGGRAPH 99). The final

More information

Development of a Computational Framework for Block-Based AMR Simulations

Development of a Computational Framework for Block-Based AMR Simulations Procedia Computer Science Volume 29, 2014, Pages 2351 2359 ICCS 2014. 14th International Conference on Computational Science Development of a Computational Framework for Block-Based AMR Simulations Hideyuki

More information

Polygonization of Implicit Surfaces

Polygonization of Implicit Surfaces Polygonization of Implicit Surfaces Hongxin Zhang and Jieqing Feng 2007-01-11 State Key Lab of CAD&CG Zhejiang University Contents Polygonization of Implicit Surfaces Other Methods for Displaying Implicit

More information

1 Proximity via Graph Spanners

1 Proximity via Graph Spanners CS273: Algorithms for Structure Handout # 11 and Motion in Biology Stanford University Tuesday, 4 May 2003 Lecture #11: 4 May 2004 Topics: Proximity via Graph Spanners Geometric Models of Molecules, I

More information

A Developer s Survey of Polygonal Simplification algorithms. CS 563 Advanced Topics in Computer Graphics Fan Wu Mar. 31, 2005

A Developer s Survey of Polygonal Simplification algorithms. CS 563 Advanced Topics in Computer Graphics Fan Wu Mar. 31, 2005 A Developer s Survey of Polygonal Simplification algorithms CS 563 Advanced Topics in Computer Graphics Fan Wu Mar. 31, 2005 Some questions to ask Why simplification? What are my models like? What matters

More information

Lecture 1.1 Introduction to Fluid Dynamics

Lecture 1.1 Introduction to Fluid Dynamics Lecture 1.1 Introduction to Fluid Dynamics 1 Introduction A thorough study of the laws of fluid mechanics is necessary to understand the fluid motion within the turbomachinery components. In this introductory

More information

L1 - Introduction. Contents. Introduction of CAD/CAM system Components of CAD/CAM systems Basic concepts of graphics programming

L1 - Introduction. Contents. Introduction of CAD/CAM system Components of CAD/CAM systems Basic concepts of graphics programming L1 - Introduction Contents Introduction of CAD/CAM system Components of CAD/CAM systems Basic concepts of graphics programming 1 Definitions Computer-Aided Design (CAD) The technology concerned with the

More information

CS337 INTRODUCTION TO COMPUTER GRAPHICS. Describing Shapes. Constructing Objects in Computer Graphics. Bin Sheng Representing Shape 9/20/16 1/15

CS337 INTRODUCTION TO COMPUTER GRAPHICS. Describing Shapes. Constructing Objects in Computer Graphics. Bin Sheng Representing Shape 9/20/16 1/15 Describing Shapes Constructing Objects in Computer Graphics 1/15 2D Object Definition (1/3) Lines and polylines: Polylines: lines drawn between ordered points A closed polyline is a polygon, a simple polygon

More information

Metafor FE Software. 2. Operator split. 4. Rezoning methods 5. Contact with friction

Metafor FE Software. 2. Operator split. 4. Rezoning methods 5. Contact with friction ALE simulations ua sus using Metafor eao 1. Introduction 2. Operator split 3. Convection schemes 4. Rezoning methods 5. Contact with friction 1 Introduction EULERIAN FORMALISM Undistorted mesh Ideal for

More information

Advective and conservative semi-lagrangian schemes on uniform and non-uniform grids

Advective and conservative semi-lagrangian schemes on uniform and non-uniform grids Advective and conservative semi-lagrangian schemes on uniform and non-uniform grids M. Mehrenberger Université de Strasbourg and Max-Planck Institut für Plasmaphysik 5 September 2013 M. Mehrenberger (UDS

More information

Simulation of Automotive Fuel Tank Sloshing using Radioss

Simulation of Automotive Fuel Tank Sloshing using Radioss Simulation of Automotive Fuel Tank Sloshing using Radioss Prashant V. Kulkarni CAE Analyst Tata Motors. Pimpri, Pune - 411018, India Sanjay S. Patil Senior Manager Tata Motors. Pimpri, Pune - 411018, India

More information

Chapter 6 Visualization Techniques for Vector Fields

Chapter 6 Visualization Techniques for Vector Fields Chapter 6 Visualization Techniques for Vector Fields 6.1 Introduction 6.2 Vector Glyphs 6.3 Particle Advection 6.4 Streamlines 6.5 Line Integral Convolution 6.6 Vector Topology 6.7 References 2006 Burkhard

More information