All-optical filtering of nuclear magnetic resonance logging data based on a diffractive neural network

Area of Science:

• Geophysics
• Data Science
• Optical Computing

Background:

• Nuclear magnetic resonance (NMR) logging data suffers from low signal-to-noise ratio (SNR), hindering accurate stratigraphic evaluation.
• Current filtering methods like U-Net and MsEDNet have limitations in parameter adjustment for logging data, leading to distorted results and low SNR.

Purpose of the Study:

• To propose a novel optical diffractive neural network (DNN)-based filtering system for enhancing NMR logging data.
• To address the challenges of low SNR and signal distortion in NMR logging data processing.

Main Methods:

• A 1D echo data is converted to 2D data using the Sinkhorn-Knopp algorithm for optical diffractive computing.
• A residue DNN is employed to effectively separate noise from the NMR logging data.
• The proposed system aims to protect signal integrity and avoid neural network degradation.

Main Results:

• The optical diffractive DNN-based filtering system achieves a higher SNR compared to U-Net and MsEDNet.
• The proposed method effectively separates noise, preserving the integrity of the NMR logging signal.
• Simulation and experimental results validate the effectiveness of the developed filtering system.

Conclusions:

• The optical diffractive DNN offers a superior solution for filtering NMR logging data, overcoming limitations of existing methods.
• This approach enhances data quality for stratigraphic evaluation by improving SNR and reducing signal distortion.