Seismic velocity modeling in the digital transformation era: a review of the role of machine learning

Abstract Seismic velocity modeling is a crucial step in seismic processing that enables the use of information from both and wells to map depth thickness subsurface layers interpreted images. The can be obtained form normal moveout (NMO) or by an inversion (optimization) process such as full-waveform (FWI). These methods have several limitations. limitations include enormous time consumption case NMO due manual heavy human involvement picking. As optimization problem, it incurs high cost suffers nonlinearity issues. Researchers proposed various machine learning (ML) techniques including unsupervised, supervised, semi-supervised model more efficiently. focus studies mostly automate picking, improve convergence FWI, apply FWI using ML directly data. In purview digital transformation roadmap petroleum industry, this paper presents chronologic review these studies, appraises progress made so far, concludes with set recommendations overcome prevailing challenges through implementation advanced methodologies. We hope work will benefit experts, young professionals, enthusiasts help push forward their research efforts achieving complete automation further enhancing performance applications used framework.