Study Of Marinediesel Generator Fault Diagnosis Based On Neural Network

The Marine power station is motive heart of the whole ships system, andwhich supplies power for the system. Gradually with the high performance and thestructure complication complex of the modern ship system, the diesel generator setwhich constitutes marine power station also increasingly develops toward large andhigh speed and precision. With the work performance continuously improving, thedegree of automating is higher. On the one hand it will rise to the rate of production,rise the electric power quantity of the ships system, lower the maintaining cost andenergy consuming; But on the other hand, the problem brought is that once some partor some a link among them go wrong, which usually will make the whole shipssystem in paralyze, so it directly or indirectly results in huge economy loss, evenresults in the key equipments damage, and endangers people safety. Therefore, how toquickly judge the reason of the fault and availably eliminate the faults, and guaranteethe ships continuing the normal voyage has special important meanings.Based on the analysis of domestic and foreign research about intelligentdiagnosis technology and fault diagnosis methods of diesel engine, a method of dieselengine fault diagnosis based on genetic neural network is given in this paper. Thefault diagnosis of fuel system is deeply studied.Firstly, the research status of fault diagnosis technology and contents of the topichave been discussed briefly in this paper; several popular methods of diesel enginefault diagnosis are introduced; the shortcoming of neural network in fault diagnosis isanalysized.Secondly，for shortcomings of slow convergence rate and falling into localminimum easily of BP neural network, genetic algorithm and neural network arecombined. Using the global search ability of genetic algorithm, initial weights andthresholds of neural network are optimized and the two inherent shortcomings ofnetwork are solved, which has enhanced the accuracy and rapidity of neural networkin fault diagnosis.Lastly, simulation test is carried on by the MATLAB, the network is constructedand trained using typical fault data of fuel system as genetic neural network training samples and simulation fault is diagnosed and analysized. The simulation resultsshow that the fault diagnosis result based on genetic neural network is well consistentwith measured values. As long as we choose enough typical initial fault samples totrain genetic neural network, the stability of network is better. The method of faultpattern recognition based on genetic neural network can fully use informationcharacteristics, achieve mapping relation between input and output and get accurateresult.