Application of Multi-Dimensional Deconvolution to Enhance Seismic-While-Drilling Data Imaging

Abstract Seismic-while-drilling (SWD) provides a cost-effective solution to sense the subsurface by utilizing the drill-bit noise as a seismic source. However, extracting useful information from SWD recordings remains challenging because the source signature is often erratic and, without a high-quality pilot signal, cannot be reliably determined. Thus, we propose a practical workflow for applying multi-dimensional deconvolution (MDD) to SWD data, producing a reflection response that has a controlled source signature and is free of surface-related multiples. Two key components of the proposed workflow are the estimation and suppression of the direct arrival, based on the particle swarm optimization algorithm, and the data preparation process for the MDD wavefields; namely, we adopt a strategy commonly used in seismic interferometry whereby auto- and cross-correlation is applied to time segments of the continuous SWD recordings, and the resulting correlograms are stacked together. MDD is then applied using such waveforms as inputs. The proposed methodology is first validated on a range of synthetic data modeled with an acoustic wave equation, demonstrating the improved quality of the retrieved virtual data and imaging results compared to those obtained from standard correlation-based interferometric redatuming. An additional synthetic test using elastic modeling further confirms the robustness and effectiveness of the proposed workflow in scenarios more representative of real field data. Finally, the proposed method is successfully applied to a field dataset. Both the synthetic and field results demonstrate the effectiveness of this workflow, offering a practical and economically viable solution for processing SWD data.