Accelerating Marine Seismic Preprocessing with Machine Learning

This technical report discusses the application of machine learning (ML) to enhance marine seismic data preprocessing, specifically focusing on reducing turnaround times in imaging projects. It outlines the challenges faced in seismic imaging, such as increased algorithmic complexity and static turnaround times despite advances in computing power. The report details traditional fast-track solutions, which, while quicker, often compromise data quality. In contrast, ML solutions are presented as a method to maintain data integrity while improving efficiency. The report describes three key preprocessing steps: swell noise removal, deghosting, and designature, utilizing deep neural networks to automate and optimize these processes. Results from a 3D project in the Niger Delta are showcased, demonstrating the effectiveness of ML in improving data quality and reducing processing times. The findings indicate that ML can significantly streamline workflows in marine seismic imaging, offering a viable alternative to conventional methods while ensuring high-quality outputs.