Design of Experiments for Aeroelastic Analysis

The performance of an aircraft in flight is, in part, a result of interactions between aerodynamic forces and structural deformations. Aerodynamic pressures result in elastic deformations which alter the wing shape and thus affect the aerodynamics. Consideration of this multidisciplinary interaction is critical to wing design. In particular, divergence (static elastic instability) and flutter (dynamic resonance) are potential catastrophic effects to be avoided. Performing an aeroelastic analysis requires the combination of static and dynamic structural analysis (often done through finite element analysis) with aerodynamic analysis (typically using some form of computational fluid dynamics, CFD). For large grids, each of these can require a significant computational effort. Resolving the interactions between the two is an iterative process which only magnifies the problem. This is a typical characteristic and drawback of multidisciplinary analysis; it makes exploring a large design space (which may include a large range of wing shape, structural support, and material choices) particularly challenging. Statistical design of experiments (DOX) is one technique for design space exploration using a limited number of targeted, computational experiments. DOX is useful for identifying the design variables most critical for a relevant response, and for finding sensitivities needed for design optimization. The objectives for this project were (1) to find the most significant geometric, modeling, and material parameters that affect the predicted aeroelastic responses of a simple wing geometry, (2) to develop parsimonious, low-order response surfaces to model effects of interest, (3) and to evaluate the quality of the response surfaces. The computational experiments were performed with MSC Nastran which combines finite element analysis for the structural response with a steady vortex-lattice method for trim aeroelastic analyses. The discussion will include an overview of the experiment design selection process, formulation of an approximation model, and an explanation of key metrics for evaluating the response surface designs. Comprehensive results are presented for the natural frequency responses, as well as a preliminary analysis of aerodynamic trim solutions.