RBF神经网络在表面粗糙度光纤传感器中的应用

Application of RBF Neural Networks in Surface Roughness Optical Fibre Sensor

本文提出了以径向基函数 (RBF)神经网络处理表面粗糙度光纤传感器输出信号的方法。将传感器的输出信号及作为光源的激光器光强信号同时加在 RBF神经网络的输入端 ,利用 RBF神经网络能够以任意精度逼近非线函数的能力的优点 ,同时实现对传感器的非线性补偿及减轻激光器输出光强变化带来的影响。采用这种方法的表面粗糙度光纤传感器 ,降低了对激光器输出功率稳定性的要求 ,具有测量范围大。

A new method for processing output of surface roughness optical fibre sensor by radial basis function (RBF) neural network is presented. The sensor's output and the laser intensity signal that used as light source in the measuring system are put on the RBF neural network inputs. Utilizing the feature of RBF neural network powerful ability for function approximation, we realize the calibration of sensor that suffers from nonlinearities. The measurement result is also not affected by laser intensity variation. This greatly reduces the laser power stability requirement. It is possible to obtain larger measuring range and higher measuring accuracy.