This investigation sought first to determine the feasibility of generating a surrogate model of the interference drag between nacelles and wing-fuselage systems suitable for the inclusion in a multidisciplinary design optimization (MDO) framework. The target aircraft was a single-aisle, transonic aircraft with a freestream Mach number of 0.8 at 35,000 feet and a design lift coefficient of 0.5. Using an MDO framework is necessary for placing the nacelle because of the competing objectives of the disciplines involved in aircraft design including structures, acoustics, and aerodynamics. A secondary goal was to determine what tools are necessary for accurately capturing interference drag effects on the system. This research used both Euler computational fluid dynamics (CFD) with a coupled viscous drag estimation tool and Reynolds Averaged Navier-Stokes (RANS) CFD to estimate the system drag. The initial trade space exploration that varied the nacelle location across a baseline airframe configuration was completed with the Euler solver, and it showed that appreciable overlap between the wing and nacelle led to large increases in interference drag. A follow-on study was conducted with RANS CFD where the wing shape was tailored for each unique nacelle position. In comparing the results of the Euler and the RANS CFD, it was determined that RANS is required to accurately capture the flow features. Euler solvers can create artifacts due to the lack of viscous effects within the model. Wing tailoring is necessary because of the sensitivity of transonic flows to geometric changes and the addition of neighboring components, such as a nacelle. The research showed that for above and aft wing locations, a nacelle can overlap the trailing edge without incurring a drag penalty. Nacelles placed in the conventional location, forward and beneath the wing, displayed low interference drag effects, as the nacelle had a small and local impact on the wing's aerodynamics. Given the high cost of computing a RANS solution with wing tailoring, and the large design space for nacelle locations, building a surrogate model for interference drag was found to be prohibitive at this time. As the cost of computing and mesh generation decreases, collecting the data for building a surrogate model may become tractable. / Doctor of Philosophy / Engine placement on an aircraft is dependent on multiple disciplines. Engine placement affects the noise of the aircraft because the wing can shield or reflect the engine noise. Engine placement impacts the structural loads of an aircraft, with some positions requiring more reinforcement that adds to the cost and weight of the aircraft. Aerodynamically, the engine placement impacts the vehicle's drag. Taken together, the only means of trading the different disciplines' needs is through a multidisciplinary design optimization (MDO) framework. The challenge of MDO frameworks is that they require numerous solutions to effectively explore the trade space. Thus, MDO frameworks employ fast, low-order tools to compute hundreds or thousands of different combinations of features. A common approach to make running MDO analysis feasible is to develop surrogate models of the key considerations. Current aerodynamic drag build-ups for aircraft do not consider the interference drag associated with engine placement. The first goal of this research was to determine the feasibility of generating a surrogate model for inclusion in an MDO framework. In order to collect the data required for the surrogate, appropriate tools to capture the interference drag are required. Building a surrogate requires a large number of samples, thus the aerodynamic solver must be fast, robust, and accurate. An Euler (inviscid) computational fluid dynamics (CFD) was used do explore the engine placement design space to test the feasibility of building the surrogate model. The target aircraft was a single-aisle, transonic aircraft with a freestream Mach number of 0.8, flying at an altitude of 35,000 feet and a design lift coefficient of 0.5. The initial vehicle used a baseline wing, and the engine placement was varied across the wing span and fuselage. The results showed that the conventional location, where the engine is forward and beneath the wing, had the a modestly beneficial interference drag, though positions near the trailing edge and above the wing also showed neutral interference drag. In general, if the engine overlapped the wing, the interference drag increased dramatically. A follow-on study used Reynolds Averaged Navier-Stokes (RANS) CFD to investigate seven engine placements above and aft of the wing. Each of these positions had the wing tailored such that the wing performance would be typical of a good transonic wing. The results showed that with wing tailoring, a moderate amount of overlap between the wing and nacelle results in reduced or neutral interference drag. This is in contrast with the baseline wing results that showed moderate overlap led to large increases in interference drag. The results from this research suggest that building a surrogate model of interference drag for transonic aircraft is not feasible given today's computational resources. In order to accurately model the interference drag, one must use a RANS CFD solver and tailor the wing. These requirements increase the cost of evaluating an engine position such that collecting enough for a surrogate model is prohibitively expensive. As computational speeds increase, and the ability to automate CFD mesh generation becomes less time intensive, the feasibility may increase. Using an Euler solver is insufficient because of the lack of viscous effects in the flow. The lack of a boundary layer leads to artifacts appearing in the flow when the nacelle and wing are in close proximity.

Strides in modern computational fluid dynamics and leaps in high-power computing have led to unprecedented capabilities for handling large aerodynamic problem. In particular, the emergence of adjoint design methods has been a break-through in the field of aerodynamic shape optimization. It enables expensive, high-dimensional optimization problems to be tackled efficiently using gradient-based methods in CFD; a task that was previously inconceivable. However, adjoint design methods are intended for gradient-based optimization; the curse of dimensionality is still very much alive when it comes to design space exploration, where gradient-free methods cannot be avoided. This research describes a novel approach for reducing dimensionality in large, computationally expensive design problems to a point where gradient-free methods become possible. This is done using an innovative application of Principal Component Analysis (PCA), where the latter is applied to the gradient distribution of the objective function; something that had not been done before. This yields a linear transformation that maps a high-dimensional problem onto an equivalent low-dimensional subspace. None of the original variables are discarded; they are simply linearly combined into a new set of variables that are fewer in number. The method is tested on a range of analytical functions, a two-dimensional staggered airfoil test problem and a three-dimensional Over-Wing Nacelle (OWN) integration problem. In all cases, the method performed as expected and was found to be cost effective, requiring only a relatively small number of samples to achieve large dimensionality reduction.