Research On Recognition Of Specific Vibration Signal Based On Distributed Optical Fiber ?-OTDR

The phase-sensitive optical time domain reflectometer is abbreviated as ?-OTDR,as a new type of distributed optical fiber sensing technology,it has considerable application prospects in many information detection fields.In various monitoring sites,simply locating the optical fiber vibration signal can no longer meet the requirements of ?-OTDR technology,which requires the signal processing method to correctly identify the type of vibration signal.In this paper,the theoretical basis and sensing principle of the ?-OTDR distributed optical fiber sensing system are studied,and a method for extracting vibration signal features combined with Hilbert transform and local mean decomposition(LMD)is proposed,and the Elman neural network is used to classify and recognize the vibration signal,which improves the accuracy of the vibration signal type recognition.The main research content of this paper is as follows:(1)The basic principle of ?-OTDR distributed optical fiber sensing system is studied and analyzed.The generation principle of back Rayleigh scattering is described,the one-dimensional mathematical model of back Rayleigh scattering signal transmission is listed,and the principle of system sensing and vibration signal detection for ?-OTDR are explained in conjunction with the principle of coherent detection.(2)A ?-OTDR distributed optical fiber vibration sensing experiment system based on coherent detection is established.The amplitude difference method was used to extract and locate the vibration events simulated by piezoelectric ceramics(PZT),and the signal-to-noise ratio of the ?-OTDR vibration sensing system is 12.7d B and the spatial resolution is 10 m.The method of two-dimensional signal matrix Fourier transform is used to realize the simultaneous extraction of vibration signal position and frequency information.(3)This paper mainly introduces the core of the vibration signal classification and recognition system,uses sliding mean filtering to perform noise reduction preprocessing on the vibration signal,uses the combined method of Hilbert transform and LMD decomposition to extract the features of the vibration signals,and Elman neural network is proposed to classify and recognize the vibration signals.An offline processing experiment was performed on part of the vibration signal data,and the feasibility of the Elman neural network classification and recognition scheme was proved by experimental analysis.(4)Based on Labview and Matlab,the software platform development of ?-OTDR distributed optical fiber vibration system was completed,and the vibration signals used for classification and recognition was collected in the laboratory.The vibration signals of 40 Hz,80Hz,100 Hz and 500 Hz were collected by using PZT simulate vibration,and the classification and recognition algorithm of Elman neural network was verified by the extracted signal samples.The experimental results show that the average accuracy of Elman neural network in the classification and recognition of vibration signals of four frequencies can reach 92.5%,after adding 15 Hz vibration signals into the original four signals and re-composing the vibration signals of five frequencies,the average accuracy of classification and recognition of Elman neural network is 87.2%,and the highest recognition rate of a single vibration signal is 98%.