Fault Diagnosis Method Of Ship Parts Based On Deep Learning

Today,with the continuous development of ship technology,shipping has gradually become the main method of water transportation,accounting for about 90% of the world’s trade and transportation.At the same time,ship safety issues have gradually attracted attention.my country’s ship technology will be in a stage of rapid growth,and the ship economy has entered the national strategic vision.With the development of computer technology and intelligent algorithms,high-tech and equipment have more extensive use space in ships,and the amount of system data generated has increased sharply.As a result,the use of technology and equipment management can no longer rely solely on traditional manual mechanisms.It is also true that the data generated by the system will also play a more direct and efficient role in equipment monitoring.With the gradual expansion of equipment data,fault diagnosis technology based on expert experience and theoretical logic has higher requirements for experience and theory.Compared with the method of using intelligent algorithms to analyze data and then diagnose equipment faults,it is easier to implement.This article aims to solve the problem of fault diagnosis of ship parts in different situations based on the current background and research technology theory.This paper starts with the needs of fault diagnosis of ship equipment parts,and firstly proposes a one-dimensional part-level fault diagnosis model.And on this basis,the practical problem is further deepened,the model is applied to small sample unbalanced data,combined with the situation that a large number of normal state data and a small amount of fault data often appear in the production operation of equipment,a new idea based on generative adversarial network is proposed.The fault diagnosis model combined with it realizes the generation of different types of fault data,which is used to solve the problem of fault diagnosis that a large amount of data is in normal state and some data is in fault state in the normal use of key components of ship equipment.This article considers the particularity of marine equipment shipping,and applies the fault diagnosis background environment to it.A fault diagnosis method of deep transfer learning for cross-domain small sample data is proposed.In the course of ship operation,due to its special navigation conditions,the background of fault diagnosis is under different working conditions.The method proposed in the paper uses the fault diagnosis model through the deep transfer learning mechanism,and applies the fault diagnosis model suitable for a certain working condition to the equipment under another working condition for fault diagnosis work,and adapts the characteristics in the transfer learning mechanism part Combining with instance adaptation,the two-dimensional control method can better realize the closer one-dimensional fault data distribution for the same equipment under different working conditions,so as to effectively solve the fault diagnosis problem.