Research On Identification Method Of Temperature Creep Model Of DTS System

Distributed optical fiber temperature sensing system is a new distributed real-time sensing temperature measurement technology based on fiber spontaneous Raman scattering effect and optical time domain reflection technology.In the research of distributed optical fiber temperature sensing system,it is found that the photoelectric device will be affected by temperature and gradually aging with time,causing the temperature correction coefficient to change,resulting in errors in the measurement output.Moreover,with the objective factors such as distance and time,optical fiber will also produce certain loss in transmission.If no improvement measures are taken,the temperature measurement will be in error,and it is difficult to meet the high-precision measurement requirements of the long-term online operation in the industrial environment.Aiming at this problem,this thesis proposes a nonlinear signal compensation method for distributed optical fiber temperature sensing system based on radial basis neural network.Firstly,for the commonly used neural network algorithms,compare and summarize,select the appropriate algorithm to build the system model.Through simulation analysis,the error back propagation neural network and radial basis function neural network are compared in network structure,network construction,training algorithm,network training,network resource utilization and network effect.It is concluded that the radial basis function neural network has strong generalization ability,fast learning speed and high approximation accuracy.The characteristics of the radial basis neural network make it more dynamic than the error back propagation neural network.Secondly,a non-linear compensation model of distributed optical fiber temperature sensing system based on radial basis function neural network is designed.Aiming at the influence of dispersion constant on radial basis function neural network,based on feedback control principle,control error automatically adjusts dispersion constant,improves model accuracy,accelerates network convergence and improves overall performance.The experimental results show that the best effect is obtained when the distribution constant is 847,and the error range is increased from-10.382%~-0.875% to-0.966%~0.322%.Compared with the polynomial fitting algorithm,the error range of the polynomial fitting model is-5.107%~4.943%.The difference between the error range of the two methods is about ten times.Therefore,the radial basis function neural network model satisfies the prediction accuracy requirement of the system.Finally,the C++ calls MATLAB engine for mixed programming to achieve the operation of the simulation platform is too cumbersome,too much memory utilization,strict experimental conditions and so on.In order to solve this problem,the output function of the trained radial basis neural network model is designed,and the non-linear compensation model of the distributed optical fiber temperature sensing system signal based on the output function is constructed.Through the layout of the interface and the compilation of the control program,the simulation platform of signal compensation for distributed optical fiber temperature sensing system is established.The simulation platform of signal compensation through text prompts can make the observation result clearer,make the viewing graphics more convenient,and realize the function of human-computer interaction.At the same time,the output function algorithm can be transplanted to other platforms,such as distributed optical fiber temperature sensing system host.