Chaotic Time Series Prediction, Analysis And Forecast In The Rotor Remaining Life

Chaos is an irregular behavior which is a wide existent phenomenon. In recent years, with the chaos has been found in fields such as meteorology, hydrology, economic and social, making the studying in application of chaos becomes one of the most significant topics currently. Chaotic Time Series Forecasting is an important research direction in forecasting research area. Rotor is the core component of a motor. When the surface cracks in the motor reach a certain width, the rotor will break. The inertia caused by higher rotation will damage the machinery equipment seriously. The research on the process of a rotor's fracture that also means the remnant life of a rotor in our country is very limited so far. So, the theoretical significance and practical value of the predicting the remnant life of a rotor using chaotic time series building model is very important. This paper does a deep research on chaotic time series forecasting and uses it in the prediction of the remnant life of a rotor.Firstly, according to the property that it can't distinguish chaotic signal form noise signal in frequency because of overlay band, in the wavelet domain, using wavelet transform to denoise. After quoting the algorithm of wavelet Spatial Correlation Filtering, this paper makes an improvement of filtering algorithm in wavelet domain and makes a simulation on Lorenz time series.The results of the simulation show that the denoising effect of the improvement algorithm is relatively good.Secondly, the paper introduces phase space reconstruction theory, makes a detailed analysis on the meaning and existing solving method of the delay time and the optimal embedding dimension,and it also do a simulation to the delay time and the optimal embedding dimension which are calculated by Using mutual information method and false neighbor method.Then, this paper has discussed the model and structure of RBF neural network. Starting from single-step predication and multi-step predication, according to features of the chaotic time series after phase space reconstruction, this paper presents four forms of chaotic time series prediction,does a simulation on Lorenz time series using RBF neural network according to the anterior four forms; makes a deep analysis in theory on the difference among simulation diagrams. The simulation shows that the multi-step indirect predication is the most effective method in the four prediction methods.Finally, after doing a noise-suppressed processing on experimental data, this paper makes a prediction on remnant life of a rotor by means of RBF neural network using multi-step indirect prediction. The result is relative ideal by using the prediction of this paper offers.