A model for predicting vibration series of turbogenerator unit is proposed by combining ANNs techniques with non-linear phase space reconstruction. Such established model doesn't make any hypothesis to history data, but reproduces the dynamic specific of attractor by phase space reconstruction. ANNs architecture to which the prediction model applys is discussed. The influences of embedding dimension m and time delay τ on prediction accuracy are also discussed in detail. Fairly well agreement can be acquired between the predictions of model and actual vibration series of unit.
展开▼
