Data-driven joint noise reduction strategy for flutter boundary prediction

Flutter test data processing is crucial for modal parameter identification, which facilitates flutter boundary prediction. However, the response signals acquired from real experiments have difficulties due to non-smoothness, multimodal mixing and low signal-to-noise ratio. A direct analysis and prediction will often lead to low accuracy on the predictions and seriously threaten flight safety. Therefore, this paper proposes a data-driven joint noise reduction strategy to improve the performance of flutter boundary prediction. Particularly, a variational mode decomposition is substantially improved by introducing an optimization algorithm. The decomposed effective signal components are reprocessed via a wavelet threshold denoising method with a soft-hard compromise threshold function. Then, based on the matrix pencil method, the modal parameters of original turbulence response signals are identified from the impulse responses generating by deep learning. The effectiveness of the presented method is verified by a comparative analysis with conventional methods.