WaveQ3D Model
Detailed Description
The WaveQ3D model computes propagation loss using hybrid Gaussian beams in spherical/time coordinates.
This model is designed to support real-time, sonar simulation/stimulation systems, in littoral environments, at active sonar frequencies. See wave_queue class for high level class diagram.
- Ray solutions to the eikonal equation are computed in latitude, longitude, and altitude coordinates to match wide-area environmental databases. This avoids the need to translate sound speed and bottom parameters into a series of 2-DxN radials. Avoiding this step results in a significant computational savings, especially when the dynamics of the scenario precludes the reuse of environment parameters over multiple runs.
- Hybrid Gaussian Beam techniques for propagation loss calculation are used to extend the applicability of ray theory to lower frequency regimes. The Gaussian Ray Bundling (GRAB) model has successfully used these techniques to support frequencies as low as 150 Hz.
- Numerical integration of the eikonal equation is performed in the time dimension to maintain the phase continuity of the wavefront. This choice speeds up the eigenray computation by allowing a single target collision to serve as the basis for multiple Gaussian beam calculations.
- This 3-D approach also supports out-of-plane reflection from the ocean bottom. Models that use 2-D environments over a series of radials cannot support this littoral effect.
There are several limiting assumptions built into the current implementation that may limit its utility for some applications.
- The ocean data package defines the interfaces between this model and representations of the synthetic natural environment. This package also includes example implementations that have been derived from public sources. Implementations of limited distribution ocean models, such as those found in the U.S. Navy's Oceanographic and Atmospheric Master Library (OAML) are expected to be supported in a package that is external to the core model.
- The gridding of launch angles and wavefront times can cause small scale features of the environment to be skipped over. In this way, the method of characteristics used by all ray theories is a spatial application of the Nyquist-Shannon sampling theorem that limits the accuracy of any digital representations of an analog processes. Developers should analyze the sensitivity of ray gridding choices to their own environmental scenarios.
- Gaussian beam computation of propagation loss does not produce valid answers in the outermost cells of the D/E and AZ grid. This is because the model needs a supporting Gaussian beam on either side to maintain proper normalization. For this reason, all ray fans should include at least 3 rays on each side of the desired azimuth.
- The model operates at its maximum efficiency when there are a large number of targets for each sensor, but the number of targets is much smaller than the number of environmental data points. For example, the model should produce valid answers for full field propagation loss estimates at all depths and ranges, but it may run significantly slower and with significantly greater memory requirements than a scenario with just ship and submarine targets.
- The accuracy of the attenuation calculation is limited by the accuracy of the path length estimate between wavefronts. The current implementation uses straight line paths between equivalent points on the wavefront. This design decision was made to speed up the computation of low and mid-frequency Navy applications, where the effect of this approximation should be small. It's impact on weapon frequency propagation modeling has not yet been evaluated.
- The UBLAS library requires gcc -03 optimization and the NDEBUG symbol defined to take full advantage of its computation speed features. The UBLAS library does not appear to produce correct answers at optimizations lower than the -O1 level.
These assumptions should be reviewed by the system development team before this model is evaluated for their intended use.