-
Notifications
You must be signed in to change notification settings - Fork 626
Smokeview Road Map
This roadmap outlines research plans for advancing the capabilities of Smokeview, software used to visualize results generated by the Fire Dynamics Simulator (FDS).
The FDS road map
describes research plans for enhancing the fire model FDS.
The Bell Labs mathematician/engineer, Richard Hamming , stated that The purpose of computing is insight not numbers. This is the primary purpose of Smokeview, to provide insight for computations performed by the fire model FDS. This is done by converting the numbers generated by FDS into images. To be most effective, FDS and Smokeview need to be developed in concert. As advances are made in FDS, corresponding advances need to be made in Smokeview so that one can continue to interpret results.
The original design strategy for FDS and Smokeview was for FDS to perform fire simulations and for Smokeview to visualize the results. This distinction became blurred with the addition of algorithms in Smokeview for visualizing 3D smoke, since this visualization technique involves the solution of the radiation transport equation (RTE). This is the same equation used by FDS to model radiative heat transfer. As a result, Smokeview now performs physics based computations to visualize smoke.
Several areas of research need to be addressed in order to improve Smokeview. First, visualization algorithms need to be improved and others need to be added in order to more effectively interpret fire modeling data. Second, wildland urban interface fires need to be visualized more effectively, exploiting the unique nature of these kinds of fires. Finally, scenarios involving large number of grid cells need to be visualized more effectively and efficiently. Results from this last area of research benefits both physics based visualization methods and WUI fire visualizations since both often require high resolution computations.
Areas of research to accomplish these goals are described in more detail below.
Realism is a metric used by Smokeview to qualitatively gauge the accuracy of the fire and smoke display. Realism itself, however, is not the primary goal. Realism is a side effect of the application of better physics. The primary goal is to gain a more complete understanding of the data. This requires a more accurate representation of the underlying data which in turn requires the use of better physics.
Smokeview presently visualizes smoke by over-laying partially transparent planes. The smoke opacities in each plane are computed using the Beer-Lambert law, smoke opacity = 1 - exp(-ksdx), using soot propagation data computed by FDS. The series of separate partially transparent planes are combined into one image using graphics hardware and conveys what is obscured and what is not due to smoke.
Some ways in which smoke/fire visualization may be improved are
-
Investigate techniques for visualizing fire using physics based methods by using information about the fire such as its temperature and composition (soot and various gas species) to color it more realistically. Flame temperatures need to be modeled directly by FDS in order to use the black body temperature to represent flame color. Further, the resolution of the computation needs to be consistent with the desired resolution of the flame image.
-
Investigate algorithms for modeling the interaction of light and smoke. The transport of light into and out of the smoke/fire, i.e., scattering, is another important factor that effects its appearance. Light sources could consist of man made lights or light from the fire itself. Work also needs to be done to generalize the wavelength of light used to visualize the scene, i.e., implementing an algorithm in Smokeview to simulate a thermal imager or infrared viewer.
WUI fire simulation cases often involve a fire with wind moving across a terrain with embedded structures. In order to to visualize WUI cases more effectively these elements need to be addressed visually. In more detail:
-
We need a visualization model for fire brands, how they move based on a wind field generated by FDS or a wind field specified by Smokeview inputs.
-
Visualization of simulation data needs to be overlayed on the map textures used to represent the terrain.
-
Velocity data needs to be visualized if recorded experimentally.
-
The unique nature of WUI fires (i.e., fire lines) needs to be exploited to produce more effective visualizations. Besides using local slice file contours to represent the fire, line contours can be used to specify various simulation quantities on the terrain.
Smokeview objects (visual component of an FDS device) are used to represent trees and can be used to represent various building structures or flow velocity vectors at experimental sensor locations. The use of theses objects can be exploited to visualize WUI cases more effectively. In more detail:
-
The Smokeview object for a tree crown needs to be enhanced by having several states that represent various conditions of the burning tree. These conditions are represented visually using different colors and structural appearance (e.g., only trunk present colored black when the crown is burned out).
-
A Smokeview object needs to be created to represent a building structure and the various ways that ignition can occur (e.g., unburned, scorched, ignited from heat flux, ignited from firebrand flux).
-
Similarly, create a smokeview object to represent vegetation with possible states: dead, live, drought stressed, etc. These states would be represented by different user-controlled colors.
The continual availability of more powerful computers allows one to simulate fire scenarios with greater resolution, i.e., more grid cells. Techniques need to be investigated for visualizing these high resolution cases more effectively. Updating Smokeview to allow 64 bit memory addressing has already been accomplished but this is just the first step. This is a brute force technique. More refined techniques are required for selecting and visualizing only the data of interest. The fire line used to visualize WUI (wildland urban interface) simulations is a good example of this. A fire line in the context of Smokeview is simply a visual display of temperature only where the fire is located. Additional techniques for visualizing large cases that need to be investigated include:
-
Investigate methods for making it easier to probe or mine data, i.e., to retrieve only data of particular interest. FDS data files are now stored sequentially. For large data sets one needs to retrieve data efficiently at a particular time and region without reading the entire data set. User-defined spatial and temporal data subsets need to be loaded efficiently in order to shorten the time required to visualize cases.
-
Investigate standardized methods for storing data more efficiently and more effectively. Can this be done better and be practical?
-
Investigate more seamless ways of handling large data sets.
-
Investigate parallelization methods for visualizing data. Smokeview presently has a limited ability to execute code in parallel using a technique known as multi-threading. For example, some obstructions are smoothed in parallel using the pthreads library. Smokeview in this respect is multi-threaded. Techniques will be investigated for performing parallel rendering of images. This may be necessary as cases get larger and larger. One technique for parallelizing the visualization is to use several video cards, each one drawing a portion of the scene.
-
Design data structures displaying geometries in the form of immersed objects and algorithms for displaying data on these objects.
-
Investigate methods for using the video card to perform scientific computations. The computational power of the video card is already being exploited by Smokeview to perform simple smoke computations. It will need to be exploited even more in order to make the proposed fire coloring and smoke lighting computations practical. Techniques are being investigated for performing the computations needed to implement the more realistic fire and smoke computations discussed earlier.