Generative model: Impulse response generated from turbulence response in flutter signal. (15th March 2022)

• Record Type: Journal Article
• Authors: Duan, Shiqiang; Zheng, Hua; Zhou, Jiangtao; Wu, Zhenglong
• Is Part Of: Mechanical systems and signal processing. Volume 167:Part A(2022)
• Publication Date: 2022-03-15
• Subjects: Generative model -- Impulse response -- Turbulence response -- Flutter signal; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2021.108562
• Languages: English
• ISSNs: 0888-3270

Abstract

Highlights: An impulse response generative model is designed to process flutter testing signal. The Tanh activation function in the output layer optimize the model more effective. The method was verified by model parameter estimation and FBP using flutter test. Abstract: Impulse response facilitates ideal modal parameter estimation in flutter tests. However, impulse excitation is costly, difficult, and risky to execute in flutter engineering tests, so it is difficult to obtain an effective impulse response for modal parameter estimation. Limitations inherent to the physical testing conditions make atmospheric turbulence a common excitation in flutter testing signal processing. Modal analysis of the structural response excited by turbulence from the flutter testing signal is crucial. Existing methods for analyzing structural responses excited by atmospheric turbulence have certain shortcomings. This paper proposes a novel impulse response generative model corresponding turbulence response based on a deep neural network. Frequency and damping ratio calculations become a modal parameter estimation problem for natural excitation relevant to the turbulence response. The turbulence response can be used to calculate the corresponding impulse response through the generative model. The modal parameters of the generated impulse response are estimated by the generative model via Matrix Pencil (MP) method. The results can then be compared with the true model parameters from simulation data. The generative model is shown to accurately generate an impulse response as evidenced by comparison against physical testing data. The impulse response can indeed be generated by the turbulence response based on the generative model. The estimated model parameters of the Stochastic Subspace Identification (SSI) are compared with the generative model via MP in terms of accuracy and computation time, which indicate better on-line monitoring/warning flutter flight test effects. Flutter boundary predictions based on physical testing further validate the engineering applicability of the proposed method.