Prediction of engine total pressure distortion in improved cascaded forward network

Abstract

In order to solve the problem of the result delay in the real-time calculation of engine inlet total pressure distortion index in flight test, the multi-step prediction of engine inlet total pressure distortion index in flight test is carried out. The average prediction error of traditional cascade forward neural network prediction model is higher than traditional autoregressive integrated moving average model. An improved algorithm is proposed. By establishing a time series dynamic level model, the time series of engine inlet total pressure distortion index is divided into low dynamic series and high dynamic series by using particle swarm optimization algorithm. The cascade forward neural network prediction model is used for training and prediction respectively. The results show that the average prediction error and maximum prediction error of the improved algorithm are reduced by 3.90%, 10.66% and 3.29% and 1.38% respectively compared with autoregressive integrated moving average model and traditional cascaded feedforward neural network.