Sensitivity Analysis Methods for Complex, Multidisciplinary Systems

ABSTRACT: The overall technical objective of this Phase II effort is to develop computational tools for computing response sensitivities of parametric multi-disciplinary air vehicle systems that exhibit nonlinear dynamic behavior for use in gradient-based optimization, smart sampling, uncertainty quantification, and risk assessment. To this end, the ZONA/MIT team will extend the 2-D ZEUS time-domain unsteady adjoint solver developed in Phase I to 3-D adjoint solver. Also, a structural adjoint method formulated in Phase I will be incorporated in ASTROS to establish a 3-D time-domain ZEUS and ASTROS coupled aero-structure adjoint system. Meanwhile, based on the frequency-domain adjoint solver developed in the ZEUS linearized Euler solver, a frequency-domain flutter sensitivity system will be developed. Both the time-domain and frequency-domain systems will be applied to complex configurations to compare their computational efficiency and accuracy. The 3-D time-domain ZEUS and ASTROS coupled aero-structure adjoint system will be used to perform uncertainty quantification and risk assessment of 3-D wings. Finally, the ZONA/MIT team will incorporate the newly developed least squares sensitivity analysis methodology into the ZEUS code. This new capability will be applied to complex systems involving chaotic aeroelastic oscillations such as a 3-D wing with freeplay and 3-D panel flutter under supersonic/transonic flight conditions. BENEFIT: With performance requirements becoming more stringent and with the need for robust, optimum and cost effective, the designs of the next generation military aircraft are most likely to be non-conventional. Numerous parameters are needed for aircraft descriptions of non-conventional. In order for early identification of critical physical behaviors of those design concepts, response sensitivities with respect to those numerous parameters are required. The proposed adjoint solver can avoid proportional cost growth in sensitivity analysis and efficiently enable rich, parametric aircraft models to be optimized. Once developed, the proposed adjoint solver can be integrated into a multi-disciplinary design analysis and optimization systems as an efficient sensitivity generator for gradient-based optimization involving numerous design parameters. Thus, the proposed adjoint solver will be an enabling technology for the long-term goal of automating aircraft design.