基于BP神经网络的光纤法珀可调滤波器的非线性校正

The Nonlinearity Correction of Fiber Fabry-Perot Tunable Filter Based on BP Neural Network

针对光纤法布里-珀罗可调滤波器的驱动电压与其透射波长呈非线性关系,本文提出了基于BP神经网络的校正方法。该方法使用驱动电压做为神经网络的输入,输出得到经校正的透射波长,并用线性拟合和二次曲线拟合的校正方法与之对比。在一组驱动电压下,三种方法得到的透射波长与光纤传感分析仪Si720读出的标准值相比,使用BP神经网络校正得到的透射波长平均误差为0.11nm,最大误差为0.19nm,比其他两种方法的平均误差分别降低了80.6%和19.7%,最大误差分别降低了83.8%和26.6%。

To rectify the nonlinear relationship between drive voltage and transmission wavelength of the fiber Fabry-Perot tunable filter,a method based on BP neural network was presented.In this method,the drive voltage was input to the BP neural network and the corrected transmission wavelength was got from the output.In order to compare them,linear and quadratic curve fitting method was discussed,too.A set of drive voltage was used as input to the three methods.Compared with the standard value reading from the optical fiber sensor analyzer Si720,the transmission wavelength obtained by the BP neural network has the average error 0.11nm and the maximum error 0.19 nm.In comparison with the other two methods,the average error decreases by 80.6% and 19.7%,and the maximum error decreases by 83.8% and 26.6%,respectively.