Seismic Signal Denoising via RRTO-Optimized VMD Combined with Spectral Kurtosis and Enhanced Wavelet Packet Thresholding

Abstract

Abstract Seismic signal denoising is a key link to improve the interpretation accuracy of seismic data, especially in the complex noise background, where random noise and coherent noise will seriously interfere with the identification and extraction of effective signals. To this end, this paper proposes a variational modal decomposition (VMD) denoising method that combines rapid extended random tree optimization (RRTO), spectral kurtosis (SK) filtering, and an improved wavelet packet thresholding (WPT) strategy, aiming to achieve high-precision and robust seismic signal processing. In the algorithm implementation, the RRTO algorithm is first used to quickly optimize the number of modes (K) and penalty factor (α) of the VMD using efficient global search, which overcomes the limitations of empirical parameter setting and significantly improves the decomposition accuracy; subsequently, the intrinsic modal function (IMF) obtained from VMD decomposition is quantified based on SK with non-Gaussian characteristics to accurately screen the dominant modes in the noise; for the identified noise-containing IMF, the wavelet packet transform (WPT) with improved hyperbolic tangent threshold function is further introduced for fine denoising to mitigate the signal distortion and pseudo-Gibbs effect. The denoising effect and effectiveness of the new method are verified by simulated signals and real seismic records in the paper.