The objective of this proposal is to develop efficient sensitivity analysis methods based on adjoint techniques for multi-disciplinary time-dependent problems. The approach is based on the discrete adjoint, which will be derived and implemented on a software component basis, with the various multidisciplinary software components being linked together through a Python interface, thus preserving modularity, and enabling the application of the adjoint approach to legacy disciplinary solvers. For the Phase 1 project, the emphasis will be on non-linear time-dependent aeroelastic problems, and the adjoint derived sensitivities will be used to drive a flutter optimization problem. The approach will be fully verifiable to machine precision by including a scripted complex variable verification capability. The overall approach is designed to be extendable in Phase 2 to more complex multidisciplinary problems involving additional disciplines such as aerothermodynamics, and to provide an eventual path forward towards higher order sensitivity methods for use in optimization and/or uncertainty quantification.
Benefit: The extension of adjoint methods to time-dependent multidisciplinary problems will enable new capabilities in computational aerospace engineering. Aeroelastic optimization through optimal selection of structural parameters and tradeoffs between aerodynamic performance and structural rigidity will be enabled. This will be of immediate use to government programs involved in the design of conventional fixed wing aircraft, rotorcraft, high altitude UAVs, and other non-conventional configurations. Commercial applications include the development of design tools for fixed and rotary wing commercial aircraft that combine in a unified manner both structural and aerodynamic design considerations.
Keywords: aeroelastic Optimization, Sensitivity Analysis, Adjoint Methods, Time-Dependent Sensitivity Methods
