Modeling and sensitivity analysis of aircraft geometry for multidisciplinary optimization problems

Abstract

A new geometry management paradigm for aircraft design utilizes Computer Aided Design (CAD) systems as the source for consistent geometry models across design phases and analysis tools. Yet various challenges inhibit the widespread application of CAD models in aircraft conceptual design because current CAD platforms are not designed for automated shape optimization. In particular, CAD models built with conventional methods can perform poorly in automated design frameworks and their associated CAD systems do not provide shape sensitivities. This thesis aims to remedy these concerns by bridging the computational geometry tools in CAD with aerospace design needs. A methodology for constructing CAD models is presented using concepts of multifidelity/multidisciplinary geometry and design motion. A formalized definition of design intent emerges from this approach that enables CAD models with parameterization flexibility, shape malleability and regeneration robustness for automated design settings. Analytic shape sensitivities are also presented to apply CAD models in gradient-based shape optimization. The parameterization and sensitivities for sketches, extrude, revolve and sweep features are given for mechanical design; shape sensitivities for B-spline curves and surfaces are also presented for airfoil and wing design. Furthermore, analytic methods modeling the sensitivity of intersection edges and nodes in a boundary representation (BRep) are given. Comparisons between analytic and finite-difference gradients show excellent agreement, however an error associated with the finite-difference gradient is found to exist if linearizing the support points of B-spline curves/surfaces and regenerating with a geometry kernel. This important outcome highlights a limitation of the finite-difference method when used on CAD models containing these entities. Finally, various example design problems are shown which highlight the application of the methods presented in the thesis. These include mechanical part design, inverse/forward design of airfoils and wings, and a multidisciplinary design space study. Gradient-based optimization is used in each design problem to compare the impact of analytic and finite-difference geometry gradients on the final designs obtained. With each of these contributions, the application of CAD-