Intelligent Prediction Technology Reasearch On Wear Trend Of Aeroengine

Aeroengine as the important component of aircraft, in pursuit of having high performance, low cost, let the operating components working in high speed, high load and high temperature, lead to operating parts are easy to cause failure due to fatigue or abrasion. It is necessary to predict wear trend of aero engine, eliminate hidden dangers in the bud as far as possible. This method is of great importance to ensure the engine operation reliably and perform its function efficiently.Firstly iron spectrum data which can reflect running state of the engine is analyzed. According to compare the common forecast models, combining characteristics of oil analysis data, RBF network model is choosed for wear tend prediction. For the issues that the nodes of input layer affect the prediction accuracy severely, C-C method that originates from chaos theory is introduced to determine the dimensions of input samples and time delay. Then orthogonal least squares algorithm is used to establish the RBF network model. As spread parameter affects the prediction accuracy, a genetic algorithm is introduced to improve RBF network model and get the best parameters. The simulation results show that, comparing with traditional model, the prediction error of GARBF network model is smaller, generalization ability is higher and it is able to reflect the wear trend of spectral data accurately.Ferrography analysis data is another important parameter which can reflect wear status of operating parts. As ferrography analysis data of the aeroengine are affected by many complex elements, so wear trend prediction accuracy is relatively low. To solve this problem, an RBF neural network variable weight combination forecasting model (RBFNN-VWCF) is proposed for wear trend prediction of aeroengine. BP network and the SVM model are carried on for combined, genetic algorithm is used to optimize parameter. The simulation results show that, RBFNN-VWCF model taking full advantage of effective information of the two models, and it can be able to reflect the wear trend of engine more objectively. The forecasting result is robust, comparing with the single model prediction accuracy of RBFNN-VWCF model to be higher and has strong practical value, provides powerful support for decision-making of the engine next step.Finally, aeroengine wear trend forecast system is developed based on VC++6.0, SQL Server2005 and Matlab7.0. The system includes user management module, data management module, database management module and trend forecast module, it is able to predict the trend of oil analysis data accurately and locate the wear parts of the engine according to the forecast result. This system has perfect functions, friendly interface and practical value.