MULTIDISCIPLINARY INVERSE DESIGN AND OPTIMIZATION (MIDO)
G. S. Dulikravich
The Pennsylvania State University, University Park, PA, USA
Designing for improved performance and life expectancy of high speed transport configurations is traditionally conducted by performing a repetitive sequence of uncoupled, single-discipline analyses involving flow field, temperature field, stress-strain field, structural dynamics, manufacturability tradeoffs and a large amount of personal designer’s experience and intuition 1289)- (295|. Since the entire aircraft system is seemingly highly coupled, it would be plausible that both analysis and design should be performed using an entirely new generation of computer codes that solve a huge system of partial differential equations governing aerodynamics, elastodynamics, heat transfer inside the structure, dynamics, manufacturing cost estimates, etc. simultaneously. Tins approach offers very stable computation since all boundary and interfacing conditions arc incorporated implicitly. On the other hand, this approach might not be the most computationally economical since different subsystems (Navicr-Stokes equations, elastodynamic equations, heat conduction equation. Maxwell’s equations, etc.) that form such a complex mathematical system have vastly different eigenvalues and consequently converge at significantly different rates to a steady slate solution. In addition, a rigorous analysis can show that even seemingly highly coupled systems arc only loosely coupled and can be analyzed scmi-sequcntially (2%). Such a scmi – scquential approach is presently used by most researchers and the industry since it can utilize most of the existing analysis and inverse design and optimization software as ready and interchangeable modules with minimum time invested in their modifications, Nevertheless, this approach is much more prone to global instability because of the often unknown and inadequately treated boundary and interface conditions.
In the remaining pan of this article the focus will he on the computational grid, acceleration of iterative algorithms and parallelization and networking issues that are pertinent to the M1DO efforts.