Aero-engine Model Correction Technology Based on Adaptive Neural Network

In this paper, a neural network-based algorithm is proposed to adapt the performance maps of engine component models for the mismatches between aero-engine simulation models and actual engine characteristics. Based on the general characteristics data of rotating components in GasTurb, a neural network capable of calculating the efficiency and mass flow of rotating components is trained. This neural network is introduced into the engine nonlinear component model to calculate the deviation between the output parameter of each section of the engine component model and the real engine performance indicators. The linear relationship between the parameters of the nonlinear model is solved by applying perturbation theory. The error between the output characteristics of the neural network and the real engine component characteristics is derived based on the simulation error, which makes the neural network is further optimized so that it can track the current real engine performance. In this paper, a model of one turboshaft engine is used as the simulation object, and the simulation of component model building and performance maps adaptation is carried out. The simulation results show that the proposed performance maps adaptation algorithm can effectively improve the accuracy of the component-level model of the turboshaft engine, and is applicable to model correction of various types of gas turbine engines.