Support vector regression-based multidisciplinary design optimization in aircraft conceptual design

Surrogate modeling plays an increasingly important role in multidisciplinary design optimization (MDO) associated with different areas of aerospace science and engineering. As a recent developed surrogate modeling method, support vector regression (SVR) has good capability of filtering numerical noise and is well suited for surrogate modeling problems with high nonlinearity. This work is focused on evaluation of SVR-based surrogate modeling method for the potential applications in aircraft conceptual design. Three numerical examples and an aerodynamic data prediction example are presented to show the accuracy of SVR for functions of varying complexity with and without numerical noises, and the key parameters of SVR model are studied. The SVR model is applied to the MDO problem of designing a general aviation airplane and good design result is obtained. The examples show that, SVR provides sufficient flexibility of switching between regression and interpolation, can filter noise and predict the functions well with a small number of samples, and is promising in aerodynamic data prediction and aircraft conceptual design.