基于改进GRNN的光纤光栅解调算法研究

Research on FBG Demodulation Algorithm Based on Improved GRNN Model

木材干燥室的干燥温度监测是木材加工厂的重要防火措施之一,FBG传感器的材料和传感特性使其可以监测干燥室内部温度而不会成为新的火灾诱因。针对FBG传感器波长解调过程中直接解调精度低的缺点,研究了一种基于柯西分布曲线拟合和SENet改进GRNN在光纤光栅传感器的解调算法。利用柯西曲线拟合和希尔伯特变换分割FBG反射曲线峰值区域,将峰值区域数据组成改进GRNN模型的训练样本和测试样本,将SENet结构嵌入GRNN模型提升GRNN模型的性能,通过PSO算法获取改进GRNN的最优光滑因子。实验结果表明,解调温度误差约为0.15℃,波长误差为1.8 pm,与直接解调法和未改进的GRNN模型相比,嵌入SENET结构的改进GRNN模型解调对比PSO-GRNN法和SE-GRNN法,性能分别提升约76%和约84%,有效提升了光纤光栅传感器的解调精度。

Drying temperature monitoring in the wood drying chamber is one of the important fire protection measures in wood processing plants,and the material and sensing properties of the FBG sensor allow it to monitor the internal temperature of the drying chamber without becoming a new fire trigger.In order to solve the shortcomings of low direct demodulation accuracy in the wavelength demodulation process of FBG sensor,a demodulation algorithm based on Cauchy distribution curve fitting and SENet improved GRNN in FBG sensor was studied.The Cauchy curve fitting and Hilbert transform were used to segment the peak region of the FBG reflection curve,the peak region data was composed of training samples and test samples for the improved GRNN model,the SENet structure was embedded into the GRNN model to improve the performance of the GRNN model,and the optimal smoothing factor for the improved GRNN was obtained through the PSO algorithm.The experimental results show that the demodulation temperature error is about 0.15°C and the wavelength error is 1.8pm,compared with the direct demodulation method and the unimproved GRNN model,the performance of the improved GRNN model embedded in the SENET structure is about 76%and 84%,respectively,which effectively improves the demodulation accuracy of the FBG sensor.