A neural network-based modeling approach for transient performance prediction of gas turbine engines

Considering the difficulties in obtaining the specific component maps and highly iterative performance requirements when analyzing the transient performance, modeling the transient process is quite a complicated task. With few but sufficient experimental dataset, this study built the dataset-driven spatial domain long short-term memory neural network model (spatial domain LSTM) to predict thrust and exhaust gas temperature for the transient process of gas turbine engines, meanwhile comparison with the results predicted by other neural network models. In addition, the transient performance parameters calculated from GasTurb13 common transient models are performed. Through the analysis of the results of other several model tests, the spatial domain LSTM model is able to accurately predict the transient parameters that the maximum errors of thrust, and exhaust gas temperature are within 4%. The spatial domain LSTM model has a fairly satisfactory prediction performance. The proposed approach of the neural network model to predict transient parameters of the microturbine engine greatly reduces the iterative requirements, and also provides a technical reference for the integrated design of aircraft-by-engine and prediction of transient mechanical stress.