基于数据融合的多通道光纤位移传感器  

Multi-channel fiber-optic displacement sensor based on data fusion

作　　者：肖韶荣[1] 张周财 黄新[2] 

机构地区：[1]南京信息工程大学物理与光电工程学院,江苏南京210044 [2]南京信息工程大学电子与信息工程学院,江苏南京210044

出　　处：《光学精密工程》2013年第11期2764-2770,共7页Optics and Precision Engineering

基　　金：科技部公益性(气象)行业专项资助项目(No.GYHY201106047);江苏省科技支撑计划资助项目(No.BE2010733)

摘　　要：基于多传感器数据融合理论,构建了一种多通道光纤位移传感器。用一根光纤作为输入通道,三根光纤作为输出通道,每两通道构建一个双通道光纤位移传感器。对不同双通道传感器输出进行数据融合处理,拓宽了传感器的测量范围。对传感器测量结果进行归一化处理,得到传感器的多条输出特性曲线;然后从获得的多条输出特性曲线中选取合适的区域作为传感器的工作区间;最后,分别采用回归分析和神经网络算法对选取的工作区间进行数据融合处理,并讨论了它们的适用范围和优缺点。实验结果表明,多通道光纤位移传感器结合适当的数据融合方法,提高了系统的测量精度和稳定性,采用径向基函数(RBF)神经网络法得到的最大相对误差小于1.0%。三个通道融合后,光纤位移传感器的动态范围扩展为双通道的1.5倍。A multi-channel fiber-optic displacement sensor was proposed based on the multi-sensor data fusion theory. By using a fiber as the input channel, and three fibers as the output one, every two channels constructed a two-channel fiber-optic displacement sensor. Then, the data from different two-channel fiber-optic sensors were fused to extend the measurement range of the sensor. Further- more, the sensor measurement results were normalized to obtain the plurality of output characteristic curves of the sensor. By selecting the appropriate area from the plurality of output characteristic curves, the operation range of the sensor was determined. Finally, the data were processed by regres- sion analysis and neural network algorithm respectively and the range of application, advantages and disadvantages of the methods were discussed. The results show that the measurement accuracy and stability have been improved. The greatest relative error by the Radial Basis Function（RBF） neural network is less than 1.0 %, and the dynamic range of three channel fiber optic displacement sensor is extended to 1.5 times that of dual channel sensors with appropriate data fusion.

关 键 词：光纤位移传感器 多通道 数据融合 回归分析 径向基函数神经网络 

分 类 号：TN253[电子电信—物理电子学]