Hysteresis and temperature drift compensation for FBG demodulation by utilizing adaptive weight least square support...

Area of Science:

• Optoelectronics
• Signal Processing
• Machine Learning

Background:

• Tunable Fabry-Perot (F-P) filters are crucial optical components.
• Hysteresis and temperature drift significantly degrade their demodulation performance.
• Accurate signal demodulation is essential for various optical sensing applications.

Purpose of the Study:

• To develop a novel method for compensating hysteresis and temperature drift in F-P filters.
• To enhance the demodulation performance and accuracy of tunable F-P filters.
• To address the limitations of existing compensation techniques.

Main Methods:

• A novel adaptive weight least square support vector regression (AWLSSVR) algorithm was proposed.
• Temperature drift estimation using a reference fiber Bragg grating (FBG) for other sensing FBGs.
• An adaptive weighting strategy with an asymmetric noise interval was employed to mitigate noise effects in training data.

Main Results:

• The AWLSSVR method achieved a reduced error of 8.7 pm when training and testing temperature modes were similar.
• When temperature modes differed, the error was further reduced to 5.4 pm, outperforming conventional LSSVR methods (errors > 10.8 pm and > 11.9 pm).
• The method demonstrated superior accuracy and robustness with noisy training samples without requiring additional hardware.

Conclusions:

• The proposed AWLSSVR method effectively compensates for hysteresis and temperature drift in F-P filters.
• This approach offers significant improvements in demodulation accuracy and robustness over existing methods.
• The technique is practical, covering the entire C-band and requiring no extra hardware.