New method of neural networks dynamic compensator and application

The three-axle flight simulator is an important device to measure the performances of the guidance head of a missile and is a complicated electrohydraulic servo system including nonlinearity, time variable factor and uncertainty. Its middle frame system often needs compensating for the values of damp ratio and inherent frequency of the system are usually very small. A new method which uses neural networks to dynamically compensate the system to improve the performance of the system is presented. Single nerve cell PID and three layers forward networks with a dynamic compensator are applied in NNC and NND separately and the S function is used in the middle layer. The simulation results show that the approach can not only improve the performance of the system better than a linear dynamic compensator, but also make good compensation in the condition that the plant is a variant based on on-line adjusting Neral Networks.