NASA
High Performance Computing
and Communications Program
Computational AeroSciences Project
Demonstration of Aeroelastic Optimization using Distributed Heterogeneous Computers
Objective: Current activity in the Framework for Interdisciplinary Design Optimization (FIDO) project is focused on demonstrating an aeroelastic optimization capability that is based on strongly coupled nonlinear inviscid CFD and FEM structural analysis codes. Aeroelastic computational solutions require iteration between the CPU-intensive CFD and FEM analysis codes to ensure that the vehicle shape being used for the aerodynamic calculations is consistent with the structural deformations caused by the aerodynamic loads.
Approach: The FIDO 3 system has been used to implement a tightly coupled aeroelastic analysis capability that uses a marching Euler CFD code (ISAAC) combined with an FEM structural analysis code (COMET). During each loop of the aeroelastic analysis, an interface code (ADVMOD) is used to update the surface grid points for both the FEM grid and the aerodynamic grid based on the latest shape definition. An aerodynamic volume grid suitable for the marching Euler code is then automatically generated using the updated surface coordinates. After the ISAAC code computes a new surface pressure distribution, the TRN3D code uses the shape functions from the FEM model to integrate the pressure distribution and compute the appropriate forces at the FEM node points for input into the structural analysis code. During any one optimization cycle, the coupled CFD and FEM analysis solutions typically require about five iterations of the aeroelastic loop to converge the structural deformations. After convergence, COMET provides an analytic sensitivity analysis capability for the structural thickness variables. The FIDO 3 system uses a gradient-based optimizer to update the structural design variables for subsequent optimization cycles, with the objective of minimizing total aircraft weight. The system also includes interactions with simple propulsion (engine deck) and performance analysis (Brequet range equation) codes to calculate the fuel weight for a specified cruise range.
Accomplishments: The figure shows the weight history for a demonstration run of the FIDO 3 system using twenty optimization cycles, during which the aircraft weight approaches a minimum level. The entire computation took 16 hours, with each of the analysis codes being run on a different workstation using the coordination and communication functions of the FIDO 3 system. The figure also illustrates the two quadratic polynomial equations that describe the inboard and outboard wing thickness distributions. A parametric definition for the wing skin-thickness distribution allows a small number of structural design variables to describe realistically the spanwise distribution of the wing structural thickness. The thickness of each FEM panel is obtained by discretizing this parametric definition.
Significance: A complex multidisciplinary application has been developed that demonstrates a CPU-intensive aeroelastic optimization process which combines a diverse set of computer codes connected together in a single computational framework that can harness the power of a distributed system of heterogeneous computers. The aeroelastic optimization capability is especially important for multipoint design applications.
Status/Plans: The current aeroelastic optimization involves only structural design variables, while the optimization problem now being developed involves aerodynamic-shape design variables as well. A "rubberized'"FEM will be employed that retains the same topology while conforming to the corresponding shape of the optimized aerodynamic surface. The CPU-intensive analysis codes will be run on a scalable parallel testbed computer to reduce significantly the overall wallclock time.
Robert P. Weston
NASA Langley Research Center
(757) 864-2149
email : R.P.Weston@larc.nasa.gov
