解调
材料科学
光学
波长
光纤
光纤布拉格光栅
变压器
光电子学
光学滤波器
采样(信号处理)
信号处理
电子工程
温度测量
光通信
波分复用
光环行器
频率调制
光纤激光器
相位调制
折射率
带通滤波器
光纤传感器
通道间距
脉冲整形
集成光学
抖动
作者
Hao Shi,Sufen Ren,Haoyang Xu,Yugen Qi,Guanjun Wang,Shuang Wang
标识
DOI:10.1109/jlt.2026.3691416
摘要
Arrayed Waveguide Grating (AWG) demodulation systems are constrained by fixed channel spacing and limited channel bandwidth, which tend to introduce demodulation blind spots within the available wavelength range, thereby restricting the continuity and accuracy of wavelength demodulation. To address this issue, this paper proposes a demodulation method for Fiber Bragg Grating (FBG) sensors that integrates equivalent dense spectral sampling based on a thermally tunable arrayed waveguide grating with a Multi-Scale Transformer Network. Multi-temperature optical intensity acquisition is employed to effectively expand the wavelength-domain feature space, thereby filling the sampling gaps between adjacent AWG channels. To resolve the complex nonlinear mapping between multi-channel intensity responses and FBG wavelengths, a Multi-Scale Transformer Network is constructed, which combines multi-scale feature modeling with a decoding architecture based on learnable query vectors to enhance wavelength prediction accuracy. In addition, a Neighbor-Constrained Mixup data augmentation strategy is introduced to improve the robustness and generalization capability of the model under small-sample conditions. Compared with the static sampling scheme, utilizing thermal tuning for equivalent dense spectral sampling significantly enhances wavelength demodulation performance without increasing hardware complexity, enabling continuous wavelength demodulation across the entire AWG operating band and reducing the demodulation error by approximately 74%, with a root mean square error (RMSE) of 1.787 pm. These results indicate that the proposed approach achieves a favorable balance between demodulation accuracy and system complexity, providing a cost-effective and high-performance solution for high-precision FBG sensing.
科研通智能强力驱动
Strongly Powered by AbleSci AI