Fault Diagnosis Of Ballast Water System Based On Improved Deep Belief Network

As the automation and intelligence of ships have become a major development trend in the field of ships,research on how to apply more scientific and intelligent fault diagnosis methods to ship-related systems has become a major hot spot at the moment.As an important auxiliary system for managing the ship’s draft and balance,the ship’s ballast water system has a large space span and contains many equipment.The environment is prone to system failures,which in turn can cause ship hull deformation and vibration.In order to ensure that the type of ballast water system can be diagnosed quickly and accurately,this article takes a container ship simulator ballast water system developed by Dalian Maritime University as the research object,analyzes the main working process of the system,and summarizes the characteristics of common failures of the system.A fault diagnosis method of ballast water system based on improved deep belief network is proposed.This thesis analyzes and studies the main components,working process and fault characteristics of the ballast water system.Aiming at the four common system faults,data monitoring points are set to collect the operating data of the system.Afterwards,in order to design the fault diagnosis method,the basic principles and specific processes of deep belief network,D-S evidence theory and principal component analysis were studied respectively.First,the fault diagnosis method based on deep belief network is deeply studied,and it is found that its accuracy for fault diagnosis of ballast water system is not high,so D-S evidence theory is introduced to optimize it.The diagnostic output of the deep belief network can be used to calculate the basic belief distribution function of the D-S evidence theory,which increases the objectivity,and the uncertainty reasoning of the D-S evidence theory can solve the problem of the lack of uncertainty expression in the deep belief network.Considering that when the ballast water system is working,there are many state variables that need to be collected and the data usually have a certain correlation.In order to further improve the efficiency and accuracy of fault diagnosis,this thesis proposes to use principal component analysis to give priority to the collection Perform fault identification and dimensionality reduction on the sample data,and then input the dimensionality reduction data into the deep belief network optimized by D-S evidence theory.Using the above fault diagnosis method,a fault diagnosis model for ballast water system is built.The basic structure of the model is as follows.First,use principal component analysis to identify faults and extract features from the data to remove redundant information in the sample,then input the reduced data into four deep belief networks,and finally use the fusion rules and decision rules of D-S evidence theory Further optimize the results of fault diagnosis.Input the collected data into the fault diagnosis model,and use MATLAB for simulation verification,which proves that the ballast water system fault diagnosis method designed in this experiment can effectively identify and diagnose the fault of the ballast water system,and has high efficiency And accuracy.