Emulator embedded neural networks for multi-fidelity conceptual design exploration of hypersonic vehicles

In this work, an emulator embedded neural network is proposed for efficient modeling of high-dimensional predictors with multi-fidelity data sources. The emulators represent prior knowledge or expertise implemented intrusively into a neural network in forms of legacy models, physics-based models, surrogate models, etc. In this research, multi-fidelity models are embedded into selected neurons of the hidden layers to accelerate neural network...