基于神经网络的调频连续波光纤传感器扫描非线性复原技术  

Nonlinear Recovering Technique Based on Neural Network for Frequency Modulated Continuous Wave Optical Fiber Sensors

作　　者：李阳[1] 冯正和[1] 龚建敏[2] 肖艳红[2] 廖延彪[2] 

机构地区：[1]清华大学电子工程系微波与数字通信国家重点实验室,北京100084 [2]清华大学电子工程系光电子实验室,北京100084

出　　处：《中国激光》2001年第3期232-236,共5页Chinese Journal of Lasers

摘　　要：分析了调频连续波光纤传感器中扫描非线性对距离测量的影响 ,并提出了一种复原技术。这种技术通过反向传播神经网络的学习来克服扫描非线性 ,可以在扫描源具有较强的非线性时获得对目标的精确估计。同时也研究了反向传播网络在线学习的问题 ,使这种方法可以适应环境的变化。Frequency-scanning non-linearity influences range-detection accuracy of optical fiber sensors. In this paper, the influence is analyzed and a recovering technique is proposed. Non-linearity is recovered through the learning process of backpropagation neural network. Simulation results show that very accurate estimation is achieved even under severe nonlinearity of the scanning source. The on-line learning of neural network is also investigated to make this method more practical.

关 键 词：调频连续波 光纤传感器 神经网络 非线性复原 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置]