Diagnosis And Prediction Research Of FBG Sensor Fault Based On Multi Pattern Recognition

Fiber grating sensor is one of the important sensing devices in the hull structural health monitoring system.It can monitor the stress of the hull structure and transmit the yield and fatigue data to the upper computer to diagnose and predict the safety of the ship's navigation.Due to long-term voyage in complex and harsh environments,hulls are often attacked by sea waves and corroded by seawater.This can degrade the performance of the sensor.Once the structural damage is accumulated to a certain extent,it will affect the transportation of goods and even threaten the safety of workers.Therefore,we need to research the fault diagnosis of the sensor.In this paper,the hull structural health monitoring system is taken as the research object.The pattern recognition technology is applied to fault diagnosis and fault prediction.The fault feature extraction method combining wavelet packet and principal component analysis is given.Then the three pattern recognition methods are used to solve the fault.The advantages and disadvantages of the problem are identified.Finally,the fault signal is predicted and bridged to achieve high accuracy and high reliability of the monitoring system.First,on the basis of reading relevant literature and practical analysis,this article reviews the technology of fiber grating sensing,analyzes the application of fiber grating sensors at home and abroad in ships,and then describes the development of fault diagnosis technology,tasks and processes.The types of failures that may occur during the ship's navigation are proposed,and the common diagnosis and prediction methods for sensor failures are analyzed.Secondly,the fault signal feature extraction of marine fiber grating strain sensor is studied.The basic principle of wavelet packet transform and its difference from wavelet transform are introduced.Then the feature extraction data after wavelet packet transform is analyzed.With more information,the Principal Component Analysis algorithm is used to reduce the dimension of the data.Finally,the fault feature data is diagnosed and forecasted.The classification and prediction results of the three common pattern recognition methods for fiber grating strain sensors are analyzed,and a fault diagnosis scheme suitable for fiber grating sensors is proposed.An algorithm combining random forest and support vector machine is designed to diagnose and predict the faults of fiber grating sensors.Finally,the effectiveness of the fault detection method is verified.