Study On Prediction By Neural Networks And IIR Filter For The Operating Parameters Of Hovercraft

For ensuring the safe operation of hovercraft, the operating parameters, such as pitch and roll, have safe and alert limits. But this threshold is determined in hydrostatic. In practice navigation, because of the wave disturbance, the operating parameters will have great volatility, frequently exceed the alert limits. Meanwhile the first-order perturbation of the wave is so close to the control-band of hovercraft that it will create a lot of unnecessary controls, waste energy, and increase the mechanical wear. For these reasons, the paper studies how to remove the hovercraft operating parameters of the wave noise.The wave data which is divided into state 2 and 3 according to the actual needs, is simulated by frequency in part. Taking into account the encounter frequency, the paper has derived an encountered wave spectrum using the encounter frequency as variables. Finally, the wave data is divided into four cases according to the different angle of wave.The digital filter has two sorts: FIR and IIR. Because the wave data does not require linear phase, the paper applies IIR filter to the filter of the wave data, and focus on the three representative IIR filter that butterworth,chebyshev II and the greatest-smooth. After comparative study, butterworth is selected for the better combined effect.To reduce the filter delay, several data have to be forecasted. Firstly, the paper applies the periodogram to the prediction of the waves. But simulation results show the precision is too poor. Finally, the paper focuses on the RBF neural network. Simulation results show that the precision is better when the next three data is predicted and the average relative error is less than 0.3. After using the prediction results to filter, simulation results show that filter delay has reduced by 2-3 points. This has proved the feasibility of the method in reducing filter delay and laid a solid foundation for future research.