A key aspect of the preliminary design process for a new generation combat aircraft is the prediction of afterbody aerodynamic drag. Current prediction methods for preliminary design are constrained in terms of number of independent geometric degrees of freedom that can be studied due to the classic circular arc or conical afterbody geometry parametrization. In addition, the amount of data available for the construction of the reliable performance correlations is too sparse. This paper presents a methodology for the generation of aerodynamic performance maps for transonic axisymmetric afterbody and exhaust systems. It uses a novel parametric geometry definition along with a compressible flow solver to conduct an extensive design space exploration. The proposed geometry parametrization is based on the Class Shape Transformation method and it enables the assessment of the aerodynamic performance of a wider range of afterbodies at the expense of one additional geometric degree of freedom. Relative to the conventional approach, this enables the exploration of a wider design space and the construction of more complete aerodynamic performance maps. This research quantifies the impact of a number of geometric degrees of freedom on the aerodynamic performance of transonic afterbody and exhaust systems at different operating conditions.

A key aspect in the preliminary design of new combat aircraft is the prediction of the afterbody and exhaust system aerodynamic drag. To meet the various operating conditions requirements for a multi-role vehicle the afterbody typically includes a variable geometry. Within the preliminary design context, this makes the aerodynamic performance prediction a difficult challenge. This research investigates reduced order models for prediction of the aerodynamic performance of axisymmetric transonic afterbody and nozzle systems for a range of aerodynamic conditions and geometric degrees of freedom. The aerodynamic performance metric of interest is afterbody drag coefficient (CD). Two reduced order models are investigated: artificial neural network and Gaussian process. The geometric variables include boattail closing angle, nozzle throat to exit area ratio and afterbody mean angle and the aerodynamic parameters are free-stream Mach number and nozzle pressure ratio. The results show that these types of reduced order models can be used for preliminary design aerodynamic performance prediction. The Gaussian process CD prediction is less accurate compared to the artificial neural network with the latter giving a prediction uncertainty of approximately ±0.01 in CD with a 2σ confidence level. The Gaussian process prediction uncertainty is approximately ±0.013 CD.