Perimeter Safety Monitoring System Of High-speed Railway Based On Optical Fiber Sensor

The safety of high-speed railway is the basic guarantee to ensure the safe operation High-speed railway is a complex long distance system with drastic climate conditions and varying geographical environment.It is of great importance to build an efficient and intelligent safety monitoring system for precise identification of potential safety hazards The fiber optic vibration sensor is a quite effective detection means with high sensitivity and simple structure,which is capable to detect any tiny vibration around very long distance environment,such that is suitable for safety monitoring of large scale infrastructures,as high-speed railway.However,the optical fiber vibration sensor suffers from some technical challenges,such as large data concurrency,complex and diverse intrusion vibration mode,and long-term stable and reliable operation.In order to improve operational reliability and achieve faster and more accurate pattern recognition of the optical fiber vibration sensing system,we first analyze the basic vibration detection principle by the optical fiber M-Z interferometry,and then we concentrate on investigating the high-speed acquisition method of wide-area multi-channel concurrent vibration signals and the fast pattern recognition algorithm.By using both software and hardware cooperative design,an environment-adaptive pattern recognition algorithm was proposed.Practical applications to the high-speed railway station validate the effectiveness of the proposed method.The key contributions of this thesis are summarized as follows:(1)A fast location and extraction method of vibration signal based on FPGA is proposed.Motivated by characteristics of intrusion vibration signals and vibration signals in silence,a fast positioning and extraction method of vibration signal is realized in a FPGA chip,and preprocess of vibration signal is also implemented in this same chip,both of which is helpful to improve the recognition rate of intrusion mode in the host computer.To achieve reliable data transmission under any abnormal conditions,the ping-pong operation of data process is designed,which guarantees high-speed data transmission and the reliable preservation of intrusion data under abnormal conditions.(2)A fast feature extraction method in time domain and frequency domain is developed.According to the characteristics of different types of intrusion events and requirements of mode classifiers,the short-time zero-crossing number,FFT transform and EMD are used to realize the hierarchical feature extraction of vibration signals,and the feature vectors which can fully reflect the vibration events are obtained,by which the computational load of the system is reduced,and the efficiency of feature extraction is improved accordingly.(3)An environment-adaptive pattern recognition method is investigated.The railway perimeters show the characteristics of difficulty to get intrusion samples,diverse intrusion modes and varying environment.A hybrid method integrated the K-MEN and the SVM are designed to realize fast and accurate classification of intrusion signals.By using accumulated effective vibration samples for different operation environments,feature vectors of vibration mode are updated continuously so that the new feature vectors converges to their true value for the specific operation environment.The optical fiber vibration sensing system developed in this work can achieve fast and accurate recognition of illegal intrusion around the high-speed railway perimeter.Its effectiveness has been verified by the monitoring system of the Foshan EMU station,which is qualified with engineering application requirements of low false alarm rate and fast and accurate pattern recognition.