Aeroengine EGT and Its Short-term Trend Prediction

Assessing the health status of aeroengines and predicting performance degradation relies heavily on exhaust gas temperature (EGT) as a key indicator. Neural network-based research techniques are commonly utilized for predicting aeroengine EGT. To circumvent the direct application of neural networks, a specialized approach focused on the characteristics of aeroengine gas path parameters is proposed for EGT and its short-term trend prediction. Initially, the engine’s gas path parameter sensor data is analyzed to reasonably partition the data using the principle of divide and conquer. Subsequently, an improved long-short-term memory neural network is introduced to capture the data’s time series characteristics, enabling accurate EGT prediction. Drawing inspiration from the growth behavior of climbing plants under the influence of supports, the EGT’s future trend is estimated analogous to a climbing plant, with "supports" such as throttle lever instructions. Numerical simulations demonstrate the method’s effectiveness in accurately predicting aeroengine EGT and its short-term trends.