| In order to ensure that the wing integrated antenna has good electrical performance during the high altitude and long endurance of the drones,it is necessary to monitor the deformation of the wing in real time.However,traditional strain gauge sensors cannot work normally in harsh environments.Therefore,Fiber Bragg Grating sensors with many advantages such as anti-electromagnetic interference,corrosion resistance,and good transmission stability are often used to monitor the deformation of the wing in real time.However,the common fiber grating demodulation equipment on the market can only perform wavelength demodulation through the microcomputer-demodulator mode,which is difficult to apply to aircraft that have strict requirements on cabin space.Therefore,it is very important to develop a set of integrated deformation and reconstruction equipment for wing,and it is necessary to develop a high-precision and high-stability fiber grating demodulation system.This thesis mainly focuses on the design and development of the wing integrated deformation reconstruction perceptron,and has carried out in-depth research on the key algorithm of fiber grating wavelength demodulation based on the tunable F-P filter,and a set of accurate and efficient demodulation signal processing procedures are proposed.The specific research content is as follows:(1)According to the characteristics of fiber grating reflection spectrum waveform data,a fiber grating signal demodulation processing flow consisting of three parts: data preprocessing,spectral peak coarse positioning and center wavelength calculation is proposed.The cause of the wavelength demodulation error is studied emphatically,the Butterworth filter is selected as the filtering algorithm,and the threshold value is 1/2 times the peak value.The network communication module is developed based on the transmission control protocol to realize the real-time communication between the upper and lower computers.The experimental verification is carried out by building a fiber grating demodulation system.The final result shows that the center wavelength deviation and the wavelength variation deviation of the proposed algorithm meet the system index requirements.(2)The peak finding algorithm is the core of the entire demodulation system.Aiming at the selection of peak finding algorithm in the demodulation process,the influence of Gaussian fitting algorithm,direct peak finding method,cubic spline interpolation algorithm on the system demodulation accuracy is deeply studied.On this basis,in view of the poor Gaussian fitting effect of some data,a peak finding algorithm based on Generalized Regression Neural Network(GRNN)is proposed,and the parameter which has a greater impact on the network calculation results,is solved by introducing Particle Swarm Optimization(PSO).The experimental results show that the peak-finding algorithm based on PSO-GRNN can effectively improve the demodulation accuracy of fiber grating sensing.(3)Since the bandwidth of the light source has a certain range,when the number of sensors is used too much,the spectrum overlap phenomenon will occur,which will cause the calculation of the center wavelength of the reflected spectrum to be inaccurate.Therefore,this thesis combines the Attention mechanism with the Gated Recurrent Unit network and proposes a new wavelength demodulation algorithm.Considering the time series characteristics of the spectral data,the gated recurrent unit network is used to perform feature learning on the spectral data.At the same time,the Attention mechanism is used to assign different weights to the input features,which makes the calculation of the prediction model more accurate.The influence of each hyperparameter of the model on the calculation results is discussed in depth.Finally,the corresponding wavelength calculation model is obtained through offline training to realize the accurate online demodulation of the overlapped spectrum. |
