Towards a Modular Framework for Aircraft Multidisciplinary Design Under Uncertainty

The effectiveness-based design (EBD) paradigm is calling for a dramatic step forward in multidisciplinary analysis and design optimization capabilities. To achieve its goal of optimizing for measures of effectiveness will require not only the incorporation of additional coupled analyses, but also the characterization and quantification of uncertainty. Synergistic application of physics models, optimization algorithms, and uncertainty quantification methods is key to attaining reasonable accuracy and computational cost. Such synergy can be enhanced with decomposition approaches to multi-disciplinary uncertainty quantification, which flexibly accommodate distributed computing and enable tailoring of the uncertainty analysis to individual discipline models. This work builds toward a decomposition-based framework for aircraft multidisciplinary design under uncertainty. The framework implements polynomial chaos expansion (PCE) methods to characterize and propagate uncertainties along with components for obtaining statistical moments and quantile estimates from PCE coefficients. To enable gradient-based optimization, each component of the framework provides analytically derived partial derivatives. The framework is applied to the following notional EBD problem, reflective of a maritime search and rescue scenario: the loiter pattern of a fixed-wing search vehicle is optimized to maximize effective tracking time with sensing and aircraft performance constraints. When uncertainties in the subject position and aerodynamics model parameters are considered, the optimal altitude and turn radius are significantly larger than those of the deterministic case. Tailoring of the uncertainty analysis to aerodynamics, propulsion, and mission disciplines allowed for reduced design and constraint variable sizes with a negligible impact on accuracy.