Research On Fault Diagnosis Method Of Marine Electric Propulsion System

In this article,the common faults in marine electric propulsion are taken as the research object.The normal and fault simulation model of marine electric propulsion system is built on the simulation platform of Maxwell,Simplorer and Simulink,and the original fault data is obtained.The fault diagnosis model is used to classify and diagnose the fault data.According to the above ideas,this article does the following research:This article takes the common faults in the ship’s electric propulsion as the research object,completes the construction of the normal and fault simulation models of the ship’s electric propulsion system and obtains the original fault data on the Maxwell,Simpler and Simulink simulation platforms,and uses the built fault diagnosis model to analyze the fault data.Fault classification diagnosis.According to the above ideas,the research work of this article is carried out from the following aspects:Firstly,the working principle of the ship’s electric propulsion system is studied.Based on this,the “Yantai-Dalian Wheel” is used as the mother ship,and it is built on Maxwell,Simplerer,and Sinulink simulation software based on its propeller parameters and permanent magnet motor parameters.Ship electric propulsion system simulation model.The simulation model is run under different working conditions,and the rationality of the model is analyzed.Secondly,in-depth analysis of the common fault mechanisms of marine electric propulsion systems,such as: open circuit of inverter switch tube,demagnetization of propulsion motor,short circuit between propulsion motor turns,short circuit of propulsion motor phase line,and propeller winding failure.Based on the previous simulation models,17 kinds of fault simulation models with different faults and different degrees are established.The fault data is used in the subsequent experiments,and the correctness of the model is analyzed using the FFT spectrum method.Then,establish a feature extraction model and study the feature quantity extraction methods commonly used in fault diagnosis: wavelet transformation and wavelet packet transformation.Due to the advantages of wavelet packet transformation in the highfrequency frequency subdivision ability,select it as a feature extraction method,and Aiming at the shortcomings of random and blind selection of wavelet packet basis,the minimum information entropy method is used for optimization.After extracting the feature vector from the optimized wavelet packet,the fault data is reduced by the LDA dimensionality reduction algorithm,and the feasibility of the feature extraction process is verified by a simple classifier.Finally,study the principle of BP neural network and support vector machine（SVM）algorithm,and build the algorithm model of both on Matlab software.By studying SVM,it can be seen that the traditional parameter selection is very blind.Select the best and compare the pros and cons of the model through the diagnosis results.