Baseline Mining Methods Of CFM56-7B Engine

Civil aviation engine is the core power system of the aircraft, its health degree directly affects the flight safety. Engine manufacturers generally adopt advanced condition monitoring system which monitors the abnormal performance trend of the engine in time, so as to avoid major faults and eliminate the hidden danger of flight safety. The engine performance is judged according to the changes of deviation of the parameter data values and the baseline values, which is the main method of condition monitoring. The core technology is the creation method of baseline model which domestic researchers still not master accurately. Therefore, the research of baselines model of engine parameters based on the CFM56-7B engine is carried out which contents three aspects:(1). State parameters data preprocessing. There are some errors of the raw data obtained through the engine state system, due to the factors such as sensor failure. The aircraft engine data was preprocessed, according to the pauta criterion, identified the abnormal points and eliminated errors.(2).Correction method for the fan speed. Fan speed correction is a key variable parameter of the baseline model engine. At present, the algorithm of the study is an ideal model which is out with the actual engineering situation, and the precision of baseline model is poor. Then the variable index fan correction model was put forward based on similarity theory. The influencing factors of index factors were analyzed by the mathematical fitting method. The function relationship between index factors and the influence factors was solved by support vector machine (SVM) and mathematical regression iteration method. Then the accurate variable index fan correction model was built. The consequence indicates that the accuracy of the method of variable index factor is higher, which provides theoretical basis for the corrected independent baseline function of engine key parameters.(3).Baseline state parameter modeling and forecasting. Firstly, the three parameters (nuclear parameters, penalty factor, and loss factor) of the Support vector machine (SVM) and the initial weights threshold of BP neural network were optimizated by genetic algorithm to further improve the prediction accuracy and general ability. Secondly, the optimized SVM and BP were respectively used to solve CFM56-7B engine baseline (Exhaust Gas Temperature, Fuel Flow, High Pressure Shaft Speed) model based on variable index of fan correction model. The test data was utilized to verify the autonomous model, validation of compared to collect real baseline values shows that the precision of model with good ability to predict the unknown data can meet the requirements by either method, the deviation trend of the parameters is solved which can accurately reflect the actual deviation trend. The model has high practical application value on the aeroengine condition monitoring.