Inverse Synthesis of Shock Time Series from Shock Response Spectra Using Ensembles of Conditional Variational Autoencoders
The research extended a conditional variational autoencoder to create an ensemble of models that synthesize candidate shock waveforms from a target spectrum. By using recurrent wavelet decoders, the method predicts multiscale wavelet coefficients sequentially to reconstruct time series. This approach outperforms previous methods on benchmark datasets, allowing for rapid generation of distinct spectral candidates.