Researchers have developed a novel method for synthesizing shock time series waveforms from their corresponding shock response spectra (SRS) using conditional variational autoencoders (CVAEs). This approach extends previous CVAE frameworks by incorporating a recurrent decoder that predicts multiscale wavelet coefficients, which are then used to reconstruct waveforms via an inverse discrete wavelet transform. An ensemble of independently trained CVAE models, each utilizing different mother wavelets and decomposition levels, was employed to generate a variety of candidate waveforms for a given SRS, addressing the limited variability of single models. Evaluations on benchmark datasets demonstrated that this wavelet-based model achieved lower spectral errors compared to earlier direct time-domain CVAE methods and a sum-of-decayed-sinusoids baseline, while also showing competitive performance against other published inverse-SRS techniques. AI
IMPACT This research introduces a more efficient method for generating synthetic shock waveforms, potentially aiding in simulations and testing across various engineering fields.
RANK_REASON The cluster contains a research paper detailing a new machine learning method for signal processing. [lever_c_demoted from research: ic=1 ai=1.0]
- alphaXiv
- arXiv
- Brake
- CatalyzeX
- conditional variational autoencoder
- DagsHub
- Gotit.pub
- Hugging Face
- IArxiv
- ScienceCast
- Sum-of-decayed-sinusoids
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
