Performance Degradation Prediction of Aircraft Auxiliary Power Unit Using the Improved SVR

The auxiliary power unit (APU) can provide compressed air and electric power for aircraft. The accurate performance degradation prediction of APU can not only provide information for condition-based maintenance, but also ensure the safety of the aircraft to a certain degree. However, due to its complexity and the stochastic working condition, it is difficult to achieve accurate prediction results by traditional time series analysis methods. To address this issue, a probabilistic prediction method named as the improved support vector regression (SVR) is proposed. Firstly, the Gaussian process regression is utilized to capture the trend features of the performance data. Then, these features are used as the input of SVR to predict the performance degradation of the APU. The improved SVR is evaluated by the real on-wing monitoring data of the APU. Compared with other two methods, experimental results show that the improved SVR obtains better prediction results.