Abstract:Machine learning-based models provide a promising way to rapidly acquire transonic swept wing flow fields but suffer from large computational costs in establishing training datasets. Here, we propose a physics-embedded transfer learning framework to efficiently train the model by leveraging the idea that a three-dimensional flow field around wings can be analyzed with two-dimensional flow fields around cross-sectional airfoils. An airfoil aerodynamics prediction model is pretrained with airfoil samples. Then, an airfoil-to-wing transfer model is fine-tuned with a few wing samples to predict three-dimensional flow fields based on two-dimensional results on each spanwise cross section. Sweep theory is embedded when determining the corresponding airfoil geometry and operating conditions, and to obtain the sectional airfoil lift coefficient, which is one of the operating conditions, the low-fidelity vortex lattice method and data-driven methods are proposed and evaluated. Compared to a nontransfer model, introducing the pretrained model reduces the error by 30%, while introducing sweep theory further reduces the error by 9%. When reducing the dataset size, less than half of the wing training samples are need to reach the same error level as the nontransfer framework, which makes establishing the model much easier.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to efficiently predict the aerodynamic characteristics of swept wings under supersonic conditions. Specifically, the paper proposes a method based on a physics - embedded transfer learning framework to reduce the high computational cost required to establish a training data set and improve the accuracy of predicting three - dimensional flow fields. The following are the main objectives of the paper: 1. **Reduce computational cost**: Although traditional machine - learning - based methods can quickly obtain the flow fields of supersonic swept wings, establishing a training data set requires a large amount of computational resources. The method proposed in the paper significantly reduces the computational cost by using the flow field data of two - dimensional airfoils to pre - train the model and then fine - tuning it with a small number of three - dimensional airfoil samples. 2. **Improve prediction accuracy**: By embedding the sweep theory, this method can consider the influence of the sweep angle when predicting three - dimensional flow fields, thereby improving the prediction accuracy. Compared with non - transfer - learning models, introducing a pre - trained model can reduce the error by 30%, and further introducing the sweep theory can reduce the error by another 9%. 3. **Reduce the number of training samples**: In the case of reducing the data set size, using this method requires less than half of the airfoil training samples to achieve the same error level as the non - transfer - learning framework, which makes it easier to build the model. 4. **Solve cross - flow effects**: Traditional two - dimensional to three - dimensional correction methods are difficult to accurately correct three - dimensional effects near the wing tip and wing root and under large - angle - of - attack conditions. This method can better capture these complex three - dimensional effects by combining physics - embedded transfer learning. ### Formula summary - **Effective two - dimensional airfoil geometry**: \[ y_{2D}(x)=y_{3D}(x)\cdot\cos(\Lambda) \] - **Effective operating conditions**: \[ Ma_{2D} = Ma_{3D}\cdot\cos(\Lambda) \] \[ Re_{2D}=Re_{3D}\cdot\cos(\Lambda) \] \[ C_{L,2D}=C_{L,3D}\cdot\cos(\Lambda) \] - **Converted cross - sectional pressure distribution**: \[ p_{3D}(x)=p_{2D}(x)\cdot\cos^{2}(\Lambda) \] - **Converted surface friction coefficient**: \[ C_{f,s.w.,3D}(x)=C_{f,s.w.,2D}(x)\cdot\cos^{2}(\Lambda) \] Through these methods and formulas, the paper aims to provide an efficient and accurate method for predicting the aerodynamics of supersonic swept wings.

Rapid aerodynamic prediction of swept wings via physics-embedded transfer learning

Transferable Machine Learning Model for the Aerodynamic Prediction of Swept Wings

Transfer Learning from Two-Dimensional Supercritical Airfoils to Three-Dimensional Transonic Swept Wings

Study of transfer learning from 2D supercritical airfoils to 3D transonic swept wings

Rapid prediction of unsteady aerodynamic characteristics of flapping wing based on GRU

Aerodynamic Optimization of a Transonic Fan Rotor by Blade Sweeping Using Adaptive Gaussian Process

Geometric-perspective transfer learning for fast aerodynamic prediction in few-shot tasks

High precision aerodynamic heat prediction method based on data augmentation and transfer learning

Unsteady aerodynamic prediction using limited samples based on transfer learning

Fast Transonic Flow Prediction Enables Efficient Aerodynamic Design

Steady-State Transonic Flowfield Prediction via Deep-Learning Framework

Prediction of transonic flow over supercritical airfoils using geometric-encoding and deep-learning strategies

Fast Aerodynamic Prediction of Airfoil with Trailing Edge Flap Based on Multi-Task Deep Learning

Machine-Learning for Nonintrusive Model Order Reduction of the Parametric Inviscid Transonic Flow past an airfoil

A Novel Hybrid Deep Learning Method for Predicting the Flow Fields of Biomimetic Flapping Wings

Deep learning-based predictive modelling of transonic flow over an aerofoil

Fast fluid–structure interaction simulation method based on deep learning flow field modeling

Deep Learning for Multi-Fidelity Aerodynamic Distribution Modeling from Experimental and Simulation Data

Fast simulation of airfoil flow field via deep neural network

An accuracy-enhanced transonic flow prediction method fusing deep learning and reduced-order model

Fast Buffet-Onset Prediction and Optimization Method Based on Pretrained Flowfield Prediction Model