Signal recognition of a Fiber Bragg grating-based orbit monitoring system based on the convolutional neural network

Abstract

In order to adapt to the diversification of train and railway technology by combining the needs of rail transit and the advantages of a grating array fiber sensing system, this paper discusses an intelligent rail transit safety monitoring system based on grating array fiber. The vibration signal sensing system based on grating array fiber can accurately locate the time and place of vibration occurrence in real time by analyzing the train vibration signal obtained by the sensing system. The possible abnormal conditions and safety risks of the train or track can be judged. At the same time, in view of the problems of low processing efficiency, low accuracy, and high possibility of omission caused by resolving and processing vibration signals only by human resources or a single feature extraction technology, this paper introduces a data analysis and processing method in a deep learning framework to analyze the collected signal directly. Deep learning is conducted through the convolutional neural network to distinguish the normal vehicle crossing data from the abnormal data, and then the abnormal data are analyzed separately. The exception category is used according to the exception data upper opportunity.