双参量偏振调制光纤传感器输出信号的分离与线性化

分析了用一个传感头实现双参量测量的偏振调制光纤传感器的工作原理，表明该传感器能同时输出两路信号，其中一路利用泡克耳斯效应测电压或利用光弹效应测压力，另一种利用旋光效应测温度。但其输出的两个测量参量之间存在交叉敏感现象，并且其输出呈明显的非线性。因此提出一种基于人工神经网络的双参量偏振调制光纤传感器输出信号分离与线性化方法。以传感器输出作为输入样本，测量参量的实际值作为输出样本，通过训练使神经网络建立传感器输出与其实际感受的测量值之间的复杂非线性关系。计算机仿真与实验结果表明，该方法不仅能在较宽的测量范围内有效地分离两个测量参量，而且能在神经网络的输出端得到理想的线性输出。

Separation and Linearization of Outputs of Polarization-Modulated Fiber-Optic Sensor for Two-Parameter Simultaneous Measurement

The operating principle of polarization modulated fiber optic sensor with a single sensing head for two parameters simultaneous measurement is analyzed. The sensor can simultaneously output two signals, which uses the Pockels effect for voltage measurement or the photo elastic effect for pressure measurement and the optical rotation effect for temperature measurement, but it suffers from the influence of the cross sensitivity and non linearity. A separation and linearization method to outputs of polarization modulated fiber optic sensor for two parameters simultaneous measurement is presented using an artificial neural network. The neural network is trained to properly represent the complex nonlinear mapping between the sensor readings and their represented outputs accurately, using the sensor readings as input data and using the actual sensed measuring values as desired output data. The results of computer simulation and experiment show that not only the two parameters can be easily separated, but also an ideal linear relationship between the sensor inputs and the neural network outputs can be obtained over a wide measuring parameter variation range.