| With the development of science and technology, the traditional maintenance model has gradually been replaced by'condition-based maintenance', which is based on the monitoring of machinery and equipment. Only by the basis of monitoring, diagnosis can avoid blindness and have specific aim.Oil spectroscopic analysis technology has been widely applied in ship building industry field. It has been one of the effective and technical means for ship machinery equipment monitoring, fault diagnosis and fault forecasting, it can effectively detect the content of abrasion resistant element in the oil, and analyze the condition of oil pollution and additives. Oil spectral data contains two aspects, on one hand, since the wearing metal components matches the materials of friction pair, so the spectral data can be taken on fault location, on the other hand, mechanical wearing condition is a process of gradual development, so the spectral data can also be taken on prediction of mechanical equipment wearing condition. The former belongs to the category of fault diagnosis, and the latter belongs to the category of condition monitoring. The specific research of paper is the latter. Oil spectral analysis technology is used for ship condition monitoring, which can find fault or fault trend soon, avoid large fault and achieve the purpose of'condition-based maintenance'. Therefore, the establishment of spectral data prediction model is of great importance.Due to the complicated condition of ship operation, the wearing element content in lubricating oil of propeller shaft and main engine is influenced by many factors, the changing trend can not be correctly predicted in traditional method. The research paper purposes a prediction method based on genetic algorithm and BP neural network for iron content in lubricating oil. What's more,6 groups of oil sample is analyzed with MATLAB,2 of them is from propeller shaft, and another 4 is from main engine. First, it need to establish the time series of oil spectral historical data, and then, prediction model is established on BP neural networks to predict the iron element content, at last, the improved BP network with GA make the average relative error within acceptable limits. Through the case analysis, the method can satisfy the needs of ship condition monitoring. |
