Project Details
Description
The fundamental task of seismic surveys is to reveal the features of the Earths subsurface using seismic reflection signals recorded by receivers located over the area of interest. In order to extract more accurate information while decreasing the survey cost, the recorded raw seismic signals must undergo a series of signal processing procedures. The existing conventional seismic signal processing methods are usually developed based on the mathematical modeling of acoustic waveform propagation. Although many feasible methods have been accordingly designed and successfully implemented, the performance is limited by the presence of the physical dynamics beyond analytical modeling. In this project, we propose to develop novel seismic deconvolution and data compression algorithms using deep machine learning techniques. This work ought to significantly enhance the seismic image quality and decrease the storage cost. Compared to the existing works, the proposed work has two novelties: 1) the attenuation compensation module will be integrated in the deconvolution neural network, which is expected to be more efficient and practical than the conventional ones. 2) the seismic data compression algorithm using deep machine learning will be trained to minimize the interpretation performance loss, which better preserves the most important geophysical characteristics for interpretation. This project is a revolutionary research direction for seismic signal processing. The expected algorithms do not only rely on a statistical model trained with data but also are enhanced by taking advantage of the analytical models. The success of this project will have huge benefits for the local oil and gas industry.
Status | Finished |
---|---|
Effective start/end date | 1/09/21 → 31/08/22 |
Fingerprint
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.