Two-Step Multifidelity Support Vector Regression in Aerospace Surrogate Modeling

Multifidelity modeling seeks to establish an efficient framework that combines mathematical models of varying complexities, providing accurate approximations while minimizing computational burden. This approach is especially appealing for balancing prediction accuracy and computational cost. We present a comparative study of machine-learning-based multifidelity analysis tailored to aerospace engineering problems, such as sonic boom prediction using the JAXA Wing–Body aircraft model and aeroelastic datasets generated parametrically using the undeflected Common Research Model-13.5 wing. A novel two-step multifidelity support vector regression is proposed and compared with existing multifidelity deep neural networks and two-step multifidelity deep neural networks from the literature. The main objective of this study is to provide an alternative approach to neural network-based multifidelity models. Three aerospace engineering problems are used to evaluate the performance of the proposed multifidelity model. Low- and high-fidelity flow solutions are obtained using the PANAIR program and SU2 suite, respectively, while ground loudness values are computed via NASA’s sBOOM code. The evaluation metrics are selected as the cross-validation score, training time, and the number of hyperparameters. The proposed two-step MFSVR method reduces the number of hyperparameters and computational cost while maintaining similar accuracy, offering an efficient alternative to traditional multifidelity deep neural network models.