Unstructured Code GUST Released in November

AeroSoft announced the release of GUST Version 1 in early November 1998. While engineers at AeroSoft are best known for their structured CFD code, GASP, nearly half of AeroSoft's research scientists specialize in generalized indexing strategies, more commonly referred to as unstructured computational fluid dynamics.

GUST represents four years of effort at AeroSoft to develop a state-of-the-art unstructured package, with all of the capabilities of GASP and more. In addition to a state-of-the-art flow solver, GUST includes two- and three-dimensional unstructured, hybrid, grid generators. In addition, like all other AeroSoft products, GUST includes an intuitive, easy-to-use GUI to facilitate problem setup.

The flow solver in GUST supports any cell type imaginable, provided the cell interfaces can be comprised of quadrilateral or triangular faces. The computational domain may also be described by an unlimited number of zones. This means GUST supports any multi-zone structured hexahedral mesh, as well as multi-zone, hybrid prismatic-tetrahedral meshes, and multizone, tetrahedral meshes. In addition, AeroSoft has released an application programming interface (API) which allows many third-party unstructured grid generation packages to directly support GUST. At the time of this writing, GUST supports structured Plot3D, ViGYAN's VGRID, ICEM CFD, Gridgen, and of course our own grid generator. With the programming library, nearly any grid currently available can be easily converted to run with GUST.

The GUST flow solver shares many libraries with GASP. These libraries include the thermodynamics, chemistry, and inviscid fluxes. This feature has allowed AeroSoft to develop advanced thermo-chemical models in the structured code and easily migrate these models to GUST. Likewise, models developed for GUST are also easily migrated to GASP.

Through this sharing of code and ideas, GUST is able to support the full set of GASP's thermo-chemical modeling, including all chemistry models and relaxation, and the full compliment of GASP's thermodynamic models. In addition, GUST utilizes the same transport property routines as GASP and supports all transport property models currently found in GASP v3.

The area where GUST and GASP differ most is in the computation of gradients for higher order spatial accuracy and the viscous fluxes. GUST implements the integral form of the full Navier-Stokes equations. In order to compute the fluxes and jacobians, gradients must be computed on the cell interfaces.

In GUST, two methods are available for this computation. Both methods are linear and therefore produce a second order accurate stencil. The first method is loosely based on Barth's linear gradient method. This method has proven to be robust and efficient for both two- and three-dimensional problems. The second method is a psuedo, k-exact method, where a polynomial of degree one approximates the gradient in the solution. Derivatives of this polynomial are trivial to compute and produce accurate gradients.

The GUST flow solver can model two- and three-dimensional turbulent flows. Wilcox's k- model is initially the only supported model for turbulent flow in GUST. This model does not require wall functions and works well for adverse pressure gradient flows. In fact, GASP was used to aid in the validation of the turbulence models in GUST.

GUST uses the anisotropy tensor approach to allow the implementation or either Boussinesq models or algebraic stress models (ASM). As turbulence models evolve, GUST is designed to allow easy implementation of new models. Implicit boundary conditions are also available in GUST which aid in the numerical stability, allowing higher CFL numbers to be used and improving convergence.

Perhaps the most exciting feature of GUST is its ability to perform distributed parallel processing very efficiently. All meshes used with the GUST flow solver are decomposed into psuedo-zones. These psuedo-zones, called partitions, are then separately assigned to different processors. The ability to breakup a domain is unlimited. Therefore, the flow solver is theoretically unlimited in the number of processors it can utilize in the solution process. Particular attention has been paid to ensure that only necessary information is passed from processor to processor, reducing communication overhead. This is accomplished by ensuring the information required for a given partition is highly localized. For example, instead of passing 5 variables across a partition boundary in order to compute a single value, the value is computed on the original partition, and then passed.


AeroSoft, Inc.
1872 Pratt Drive, Suite 1275
Blacksburg, VA 24060
(540) 557-1900
Fax (540) 557-1919
Email
questions@aerosft.com