A new on-board turbofan engine modeling method based on a BN-MGD-DNN (batch normalize mini-batch gradient descent deep neural network) is proposed. This new method adopts BN algorithm, which accelerates the network training speed and overcomes the gradient vanish problem. Hence, using the BN algorithm, the neural network adopts deeper structure, which means the network has stronger represent capacity. This Mini-batch Gradient Descent (MGD-NN) algorithm that consumes much less time to update NN parameters is adopted. Therefore, it is more suitable for training big data set and establishing a high accuracy engine model in a large flight envelope. Finally, to verify whether the proposed method could be applied to larger flight envelop, the conventional Neural Network (NN) also adopts MGD (called MGD-NN). The turbofan engine models based on these two modeling methods are both conducted within a sub-sonic cruise envelop. The simulation results show that the proposed modeling method has much higher accuracy than the MGD-NN. Moreover, the proposed method has the characteristics of less data storage, low computation complexity and good real time performance, which are the most importance index for model realize on-board.

A novel aero-engine control method based on Deep Reinforcement Learning (DRL) is proposed to improve the engine response ability. The Q-learning that is model free and can be performed online is adopted. For improving the learning capacity of DRL, the On Line Sliding Window Deep Neural Network (OL-SW-DNN) is proposed and adopted to estimate the action value function. The OL-SW-DNN selects the nearest point data with certain length as training data and is insensitivity to the noise. Finally, the comparison simulations of the proposed method with the Proportion-Integration-Differentiation (PID) that is the most commonly used as engine controller algorithm in industry are conducted to verify the validity of the proposed method. The results show that, compared with the PID, the acceleration time of the proposed method decreased by 1.525 seconds under the premise of satisfying all engine limits.

A novel modeling method, which is based on a min-batch gradient descent neural network (MGD NN), is proposed to establish an adaptive dynamic model of a turbofan engine in a large flight envelope. For establishing a high precision engine dynamic model in a large flight envelope, it always needs a very big training data. This proposed method adopts the MGD algorithm, which is more suitable to train a neural network for big training data due to it consumes much less time to update NN parameters. Dramatically, the huger training data of the MGD NN is the better generalization performance it would be. Furthermore, a regularization strategy, which will also improve the generalization performance of the MGD NN, is applied here. Finally, compared with a popular support vector regression (SVR) modeling method, the proposed method for the adaptive dynamic model of the turbofan engine is validated within a supersonic cruise envelops. The results show that the proposed method has not only much higher precision, but also less data storage and better real-time ability than the SVR method.

The research on modeling and installation performance control of three ducts separate exhaust variable cycle engine has been studied. In this paper, we established the fan, the low pressure and high pressure compressor model based on the stage-stacking method. On this basis, a component-level mathematical model is constructed and the compressor model of adjustable mode selector valve area, the third bypass and its exhaust nozzle mode are established. Besides, the research took the inlet inside and outside characteristic model into consideration. The simulation showed that the engine could be able to modulate the third bypass airflow adaptively by control of fan vane, mode selector valve area, third bypass, and main exhaust nozzle area. It is possible to continuously match the engine’s airflow demand to the inlet’s airflow supply. The results show that compared with the mixed exhaust three ducts variable cycle engine, the inlet spillage drag of the separate exhaust variable cycle engine is reduced by about 18%. Under the condition of supersonic flight, the constant flow control can significantly improve the economy of the engine. Compared with the mixed flow variable cycle engine, the inlet spillage drag of the three ducts separate exhaust variable cycle engine is reduced by about 20%.