DDSSnet: a fast strain demodulation approach for OFDR-based fiber shape reconstruction

Area of Science:

• Optoelectronics
• Fiber optics
• Machine learning

Background:

• Optical fiber shape sensing faces challenges in balancing accuracy and real-time reconstruction.
• Existing methods like cross-correlation algorithms can be slow for complex shape analysis.

Purpose of the Study:

• To develop a fast strain demodulation algorithm for optical frequency domain reflectometry (OFDR) shape sensing.
• To improve the speed and accuracy of real-time shape reconstruction in optical fiber sensing systems.

Main Methods:

• A novel fast strain demodulation algorithm, deviation calculation and deviation denoising for shape-sensing convolutional neural network (DDSSnet), was developed.
• The algorithm calibrates sensor wavelengths and compensates for phase noise.
• Shape reconstruction was performed using the rotation-minimum frame.

Main Results:

• The DDSSnet algorithm increased processing speed by 9.691 times for axial strain distribution and 9.4 times for shape-sensing results compared to cross-correlation.
• Reconstructed shapes for a cylinder and a configuration showed low maximum relative errors (0.581% and 1.170%) and average relative errors (0.204% and 0.380%).
• The proposed method demonstrated slightly lower errors than the cross-correlation algorithm.

Conclusions:

• The developed fast strain demodulation algorithm significantly enhances the speed of optical fiber shape sensing.
• The method achieves high accuracy in real-time shape reconstruction.
• This approach shows considerable promise for practical applications in various fields requiring precise shape monitoring.