Abstract

Seismic inversion is a key tool in reservoir characterization, traditionally relying on physicsbased, wavelet-dependent methods focused on amplitude analysis. Recent advancements in seismic attribute analysis and machine learning are enabling a transformative shift in how inversion is performed. This study presents a deep learning-based seismic inversion workflow that uses Convolutional Neural Networks (CNNs) to integrate multiple seismic attributes and predict key elastic properties such as Pimpedance, S-impedance, and density. Applied to four areas across the Western Offshore and Onshore Basins, the workflow consistently produced results that matched or exceeded those from conventional methods.