A physics-informed dual-head neural network for demodulation of fiber-optic Fabry–Perot sensors to avoid mode jumping

Shun Wan

doi:10.1016/j.measurement.2026.122697

A physics-informed dual-head neural network for demodulation of fiber-optic Fabry–Perot sensors to avoid mode jumping

2026-08-10

0

PRE

AI

S

Shun Wan

S

Shuaimin Wu

K

Kaibo Gao

J

Jia Liu

Q

Qianyu Ren

G

Gensen Yang

P

Pinggang Jia *

DOI：10.1016/j.measurement.2026.122697

Original

Original request for help

Save

Abstract

En 中文

• A physics-informed dual-head CNN is proposed for FP sensor demodulation. • FSR prior guides a segmented mask to mitigate local spectral overlap errors. • Cavity length prediction is decoupled into order and phase sub-tasks. • The method avoids mode-jumping even when the signal-to-noise ratio drops to 10 dB. • Tests show a 0.764 nm resolution and ±0.0654% F.S. repeatability error.

Keywords:

Fiber-optic Fabry-Pérot sensor

Spectral demodulation

Physics-informed neural network

Mode-jumping suppression

Deep learning

A physics-informed dual-head neural network for demodulation of fiber-optic Fabry–Perot sensors to avoid mode jumping

Abstract

Journal

Organization

Cited Papers

Citing Papers