Hanyuan Peng

Title: Hanyuan Peng and Seismic Data Processing Innovation.

Introduction

Hanyuan Peng is an inventor based in Paris, France, associated with Cgg Services SAS. His patent work addresses the processing of seismic data to create clearer images of underground formations.

Latest Patents

Peng is associated with 1 patent, titled “Seismic Data Processing Using DUnet.” The invention describes a DUnet engine that processes seismic data through a contractive path, an expansive path, and a model level. The contractive path extracts a code from the input data, while the expansive path uses that code and features passed through skip connections to reconstruct information. The model level combines outputs to produce a processed image and, in some implementations, an image showing the difference between the processed result and the original seismic data.

The patent also describes selecting training data through an anchor method that expands an initial subset based on similarity measurements. A reweighting layer may combine inputs from different model layers to help preserve signal-amplitude trends.

Career Highlights

Peng’s listed patent reflects work at the intersection of machine learning and geophysical data processing. The DUnet approach is designed to support the interpretation of seismic measurements used to image underground formations.

Collaborations

Peng is associated with Cgg Services SAS. His listed coworkers include Jeremie Messud and Céline Lacombe.

Conclusion

Hanyuan Peng’s patent describes a neural-network-based approach to seismic image processing, combining multistage data analysis, similarity-guided training selection, and signal-amplitude preservation.