Research On Key Performance Parameter Prediction Method Of APU

Auxiliary Power Unit(APU)is an important functional subsystem on aircraft.It provides power and gas source for air conditioning and engine startup.A small amount of APU can also provide additional thrust for aircraft.Therefore,the key performance parameter that characterizes the trend of APU performance prediction research is of great significance,which can ensure the safe operation of aircraft and save maintenance economic costs.First,Support Vector Machine(SVM)and neural network prediction methods applied to complex nonlinear systems are analyzed,and APU performance parameter data preprocessing is studied,including filling of missing values,and box-plot analysis method is to detect abnormal values of each performance parameter.The average value correction method is applied to process abnormal values,and the min-max standardization method is used to normalize APU parameter data.At the same time,JMP data analysis software is used to analyze the correlation of other parameters,and the input parameters that have strong correlation with Exhaust Gas Temperature(EGT)are selected to provide a data basis for the subsequent prediction of APU performance parameter.Secondly,in order to improve the accuracy of APU key performance parameter prediction,in view of the parameter selection problem encountered in the actual use of SVM,mutation operation is introduced,inertia weight and learning factor parameters are dynamically adjusted,and adaptive mutation PSO algorithm is used to optimize the selection of SVM penalty parameters and kernel parameters.A prediction model of APU performance parameters based on adaptive mutation PSO algorithm to optimize SVM is presented,and comparison experiments with multiple prediction models are carried out.The results show that the accuracy and stability of the proposed model are better than those of the comparison model.Finally,in view of the fact that the performance parameter prediction method of APU based on traditional machine learning cannot make full use of the time series and non-linear characteristics between parameter data,an APU performance parameter prediction method based on the combination of CNN-LSTM-Attention is proposed.A one-dimensional CNN is introduced to obtain abstract features of different attributes.LSTM network is used to memorize these features,and combine the Attention mechanism that assigns different weights to the feature states to achieve parameter prediction.The different parameter data of APU are used to predict future EGT that are not synchronized.And compared with the proposed adaptive mutation PSO-SVM model.The experimental results show that CNN-LSTM-Attention prediction model is better than the adaptive mutation PSO-SVM prediction model for asynchronous long EGT prediction,which reflects the superiority of APU performance parameter data in use of deep learning methods for prediction,and provides a certain reference for short-term APU performance change trend prediction.