Reasearch On Optical Fiber Vibration Signal Recognition Algorithm Based On Wavelet Neural Network

Nowadays,pipeline transportation has become the main way of transportation of oil and natural gas in our country,followed by the occurrence of pipeline damage caused by human factors,which makes people suffer huge losses in production and life.Therefore,it is necessary to design in good real-time,sensitive and stable optical fiber pre-warning products.Optical fiber pre-warning system is designed to solve the problem.With the advantages of large transport volume,small land area,construction cycle and low cost,the system has attracted more and more scholars' attention.A stochastic neural network named RVFL based on the principle of phase sensitive optical time domain reflectance(?-OTDR)is proposed to identify the external intrusion signal effectively.This paper first introduces the principle of ?-OTDR system and the corresponding hardware system design,including the design of optical signal generation and recovery system and distributed network node system under the demand of long-distance real-time detection.the design of ATCA chassis architecture and the signal transduction mechanism of each card in the chassis is also introduced in this paper.Secondly,the paper proposes a real-time detection and processing flow based on CFAR detection algorithm.By analyzing the characteristics of the background signal collected by the system,the paper puts forward the standardized processing for the collected signal.The corresponding filter is designed to complete the signal filtering,and finally the signal detection process is realized.Intrusion signal recognition is very important to the system.In this paper,the Hilbert transform is used to extract the envelope of the intrusion signal,then the correlation algorithm is used to further suppress the noise of the envelope signal.Finally,the multi-level wavelet decomposition is used for the multi-scale analysis of the signal.And the signal feature extraction is realized.Finally,a random neural network is proposed to train and test the intrusion signals.The network is very suitable for the optical fiber pre-warning system with long distance and large amount of data.Through the actual data test,the algorithm proposed in this paper has a higher signal recognition rate for the intrusion signal,so the algorithm has a higher application value.