An Improved Chaos Prediction Method And Its Application In The Thruster Motor System Of Underwater Robots

The thruster motor system of underwater robot is time variation system that is multivariable and close coupling and nonlinear. Under a certain parameter and running conditions, the system will arise the chaos motion such as torque serious vibrations and electromagetism noise etc, which must make the instability and vibration and un-controlled. If the reasons are not found out, the robot can not be reliable work, which occurs severely the accident and huge economic losses and badly social influence. In order to assure the system stable and secure, so it is necessary to analyze the chaos phenomenon and control the chaos with an effective method.The method combined wavelet neural networks with phase space reconstruction is proposed. The method is more different than only using wavelet neural networks, one is the problem of neural networks input node is settled, and the other is the accurate and reliable data sample is offered for wavelet neural network input by phase space reconstruction. On the other hand, through using wavelet decomposition and reconstruction, the origianl series time is decomposed into differnet frenquency range, whose fluctuation rule can be easily grasped. This method increases the neural network prediction precision, and makes it possible to select different parameters model according to predict signals.Fistly the theory and method of the phase space reconstruction are discussed, and the advanstages and disadvanstages of delay time and correlation dimension are analysed; Secondly the merits and faults of various predict methods that are used extensively at present are analysed, and wavelet analysis technology with local performance of time-freqency and multiresolution function and neural neworks with self-study and self-tuning and BP algorithm are proposed. In the meantime, the method of the phase space reconstruction and wavelet neural networks are applied to Lorenz system, the prediction is satisfied.Finally the proposed method is applied to the motor system of underwater robots. And the speed time-series of no-load condition is reconstructed. The optimal correlation dimension and delay time are determined; it is analysed with wavelet decompostion and reconstruction and the neural networks model is established in order to realize the speed prediction of thruster motor system of underwater robots. The results show that this method is effective and reduces largely the disadvantages using only BP networks. This method can be widely used in some nonlinear time series prediction and have a engineering value.