A Hybrid of NARX and Moving Average Structures for Exhaust Gas Temperature Prediction of Gas Turbine Engines

Aiming at engine health management, a novel hybrid prediction method is proposed for exhaust gas temperature (EGT) of turbine engines. This model combines nonlinear autoregressive with exogenous input (NARX) and moving average (MA) model. A feature attention mechanism-enhanced long short-term memory network (FAE-LSTM) first developed to construct the NARX model, which used identifying aircraft using condition parameters path measurement that correlate EGT. vanilla LSTM then constructing MA improving difference between actual EGT predicted given by The evaluated real flight process data compared several dynamic techniques. results show our reduces RMSE MAE least 13.23% 18.47%, respectively. FAE-LSTM can effectively deal data. Overall, present work demonstrates promising performance provides positive guide predicting parameters.