The System Stability Criterion Based On Neural Network And Its Application In FBP

Being as a modeling theory, the Artificial Neural Network (ANN) has been widely used with a lot of special advantages, such as, performance of distribution processing and self-learning. Since a system (or object) can be described by ANN model, the information of system stability would be contained. How to extract the information from the model will be focused in this thesis.Based on BP (Back Propagation) Networks, ANN model and principle are investigated in the paper. Combined the Jury criterion in control field, a new method of system stability determinant is proposed from the aspect of mathematics, in which the weights of ANN will be direct used to construct the criterion. Comparing with the modeling method using time series analysis, the rationality and validity of the proposed method is tested by using numerical simulation.Focused on the features and requirements of flutter boundary prediction (FBP), the method presented here is introduced and investigated around precision, anti-noise and short sample. The test data of wind-tunnel and flight are used to examine the application characteristics. The result shows that the method is feasible and reliable for actual engineering.