Research On Prediction Of Aircraft APU Malfunction Trend

Auxiliary Power Units(APU)is an important part of the aircraft,which can provide power and air for the aircraft and prevent air parking,which is of great significance to ensuring the comfort and safety of passengers in the cabin.First,according to the working principle of the APU,important performance parameters related to APU malfunction are determined.Among them,the exhaust over-temperature malfunction related to the exhaust gas temperature parameter and the oil temperature over-temperature malfunction related to the oil temperature parameter are mainly studied.After that,the APU performance parameter data is preprocessing,including data normalization and outlier detection.Secondly,two prediction models are proposed to predict the future values of APU exhaust gas temperature parameter and oil temperature parameter,and then analyze the malfunction trend corresponding to APU.The first model is based on Long Short Term Memory in the prediction of APU exhaust gas temperature,and uses the Correlation Matrix to select the speed parameter that has the strongest correlation with the exhaust gas temperature.Combined with exhaust gas temperature data input to a Bidirectional Long Short Term Memory model to predict future values and analyze exhaust gas temperature over-temperature malfunction trends.The second model is based on the extreme gradient boosting model in the prediction of the exhaust gas temperature of the APU,and the Affinity Propagation optimized by the Ant Lion Optimizer algorithm.The clustering section clusters the sample set,and then inputs it into an Extreme Gradient Boosting model to predict the future value of exhaust gas temperature and analyze the exhaust gas temperature over-temperature malfunction trend.Then,the model is used to predict the future value of APU oil temperature and analyze the slip.Oil temperature over-temperature trend to verify the universality of the model.Finally,simulation experiments are conducted on the two models and the APU malfunction trend is analyzed.The simulation experiment results show that the sample set APU has no exhaust gas temperature over-temperature and oil temperature over-temperature failure malfunction in the next seven flights.Quantitative analysis of the model through the prediction accuracy evaluation index shows that the training rate and prediction accuracy of both models are better than the traditional prediction models.Among them,the Affinity Propagation-Extreme Gradient Boosting model optimized based on the Ant Lion Optimizerhas high prediction accuracy than the Bidirectional Long Short Term Memory network model based on the Correlation Matrix,and the sample training time is shorter.The two prediction models proposed in this paper can better predict the main performance parameters of the APU and analyze the related malfunction trends at the same time,which is of great significance for the APU condition-based maintenance.