NASA
High Performance Computing
and Communications Program

Computational AeroSciences Project

CONCURRENT MULTIPLE AEROEALSTICITY ANALYSES USING ENSAERO-MPI

OBJECTIVE:

Accurate prediction of aeroelastic loads and the resultant structural deformations of complex geometries using the Navier-Stokes equations

Concurrent multiple aeroelasticity analyses to exploit the architecture of the distributed memory, multiple-input, multiple-data parallel computers.

APPROACH

A discipline domain decomposition approach is used for parallelization. Each discipline of the fluid, structure, and control domains is distributed onto a different group of computational nodes. Only a single node is assigned to the structure and control groups since their computational load is relatively small compared to the fluids. The fluid domain is further parallelized based on the multi-zonal method. Furthermore, the procedure is extended for concurrent multiple simulations to exploit the architecture of the MIMD parallel computers.

A more elaborate method for fluid-structure interface is also developed for the fluid-structure coupling analysis for complex geometries. A triangular finite element concept is used for the fluid-structure interface.

For a multidisciplinary application environment, MPIRUN developed at NASA Ames is used to load different applications and run them simultaneously.

ACCOMPLISHMENTS

The capability for concurrent multiple aeroelasticity analyses has been successfully tested using a Boeing SST model (arrow wing-body configuration).

Fluid-structure coupling and fluid-structure-control surface coupling simulations are demonstrated for the Boeing SST model.

SIGNIFICANCE/IMPACT

The parallel implementation of concurrent multiple analyses into the aeroelastic computational procedure enables to study aeroelasticity of complex geometries with reduced turn-around time for the flow computations in addition to the accurate prediction of aeroelastic loads involving flexible structures and control surface deflections. This technology is being demonstrated for design oriented aeroelastic computations on HSCT 1122 configuration in collaboration with Boeing.

STATUS/PLAN

The current direct coupling process assumes that each zone of the fluid grids resides on a single SP2 node. In the future, with parallelization of the flow solver, a single zone of the grid can be distributed on a large number of nodes. This will improve load balance between computational nodes and cut down the computational time further for the flow computation. More complex geometries will be used for the validation of the code.

POINTS OF CONTACT

Dr. Chansup Byun (MCAT)
Dr. Guru P. Guruswamy (NASA)
Ames Research Center
byun@nas.nasa.gov
guru@nas.nasa.gov
415-604-4526
415-604-6329