Spatio-temporal graph convolutional autoencoder for transonic wing pressure distribution forecasting
This study presents a framework for predicting unsteady transonic wing pressure distributions due to pitch and plunge movement, integrating an autoencoder architecture with graph convolutional networks and graph-based temporal layers to model time dependencies. The framework compresses high-dimensio...
Gespeichert in:
| Veröffentlicht in: | Aerospace science and technology Jg. 165; S. 110516 |
|---|---|
| Hauptverfasser: | , , , |
| Format: | Journal Article |
| Sprache: | Englisch |
| Veröffentlicht: |
Elsevier Masson SAS
01.10.2025
|
| ISSN: | 1270-9638 |
| Online-Zugang: | Volltext |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| Zusammenfassung: | This study presents a framework for predicting unsteady transonic wing pressure distributions due to pitch and plunge movement, integrating an autoencoder architecture with graph convolutional networks and graph-based temporal layers to model time dependencies. The framework compresses high-dimensional pressure distribution data into a lower-dimensional latent space using an autoencoder, ensuring efficient data representation while preserving essential features. Within this latent space, graph-based temporal layers are employed to predict future wing pressures based on past data, effectively capturing temporal dependencies and improving predictive accuracy. Four different temporal schemes have been tested, where the spatio-temporal graph convolutional network achieved the best accuracy thanks to convolution in both time and space. This combined approach leverages the strengths of autoencoders for dimensionality reduction, graph convolutional networks for handling unstructured grid data, and temporal layers for modeling time-based sequences. To benchmark the efficacy of the framework, a comparison with the Dynamic Mode Decomposition with control technique is performed. Validation is conducted using the Benchmark Super Critical Wing test case at Mach 0.74, demonstrating that the proposed approach achieves accuracy comparable to high-fidelity computational fluid dynamics simulations while significantly reducing prediction time. This work underscores the potential of the developed framework as a scalable, efficient, and robust solution for the analysis of nonlinear unsteady aerodynamic phenomena.
•Predicts unsteady pressure in transonic flow using graph-based autoencoder models.•Assesses suitability of GRU, LSTM, Attention, and STGCN for time-series modeling.•Captures shock behavior and its nonlinear dynamics from wing motion.•Provides better generalization than conventional approaches such as DMD.•Adjusts predictions according to the elliptic and hyperbolic nature of the flow. |
|---|---|
| ISSN: | 1270-9638 |
| DOI: | 10.1016/j.ast.2025.110516 |