解调
微电子机械系统
压力传感器
电子工程
信号处理
融合
材料科学
计算机科学
频率调制
传感器融合
压力测量
工程类
声学
振动
噪音(视频)
稳健性(进化)
电气工程
干涉测量
频率响应
加速度计
信噪比(成像)
工作(物理)
集成光学
作者
Haoyang Xu,Mei Sang,Tianxiang Sun,Xin Chen,Hongyu Liu,Xuesong Xiang,Shengchao Chen,Sufen Ren,Guanjun Wang,Junfeng Jiang,Tiegen Liu,Xun Yu,Shuang Wang
标识
DOI:10.1109/jlt.2026.3725101
摘要
At extreme high temperatures, blackbody radiation and multimode fringe-contrast degradation collectively collapse the signal-to-noise ratio (SNR) of sapphire-fiber MEMS Fabry-Pérot pressure sensors, inducing interference-order jumps that foil reliable demodulation. To address these challenges, we propose a physics-informed multi-domain fusion network (PhysFusion), which enables stable optical-path-difference (OPD) demodulation under low-SNR conditions (below 10 dB) while suppressing jump errors. It fuses GAF-encoded spectral morphology with FFT-domain coarse OPD features, enabling distortion-tolerant spectral learning under physically constrained interference-order localization. At 1200$^{\circ }$C, PhysFusion achieved an R2of 0.9951 and an RMSE of 0.0892$\mathrm{\mu }$m across a pressure range of 0.5 kPa to 3.5 MPa. At the same temperature, interference order discrimination accuracy reached 0.9479, with adaptive parameters used to suppress the impact of adjacent order jump errors on demodulation. These results indicate that PhysFusion preserves physical order consistency and supports stable high-temperature pressure demodulation under severe spectral degradation.
科研通智能强力驱动
Strongly Powered by AbleSci AI