Bp Neural Network-based Adaptive Inverse Control

A sort of neural network inverse control structure with PID feedbacks based on BP algorithm has been put forward on the basis of studying the development and current situation of self-adaptation in this dissertation, which combined with the existing problem of self-adaptation inverse control.With the several years' development, research of self-adaptation inverse control has been acquired plenty of progress. However, the study of self-adaptation inverse control has been given priority to linear system, which has been little to non-linear system. Firstly, analyzed the characteristic of inverse target's modeling method based on linear wave filter, which is quite successful for studying the linear system. However, it is not very much ideal to non-linear system. Especially, most systems can't be analyzed or be non-linear, its inverse model is difficult to set up by way of tradition modeling. A sort of neural network control structure has been put forward here, because of collateral structure of neural network and its great superiority displaying among the automatic controlled field in real-time. Secondly, has studied the modeling way of the neural network and back-propagation algorithm of neural network on the basis of analyzing the operation principle and characteristic of the new control structure. The reversibility and the stability of neural network inverse system has been analyzed, which have been ready for following simulation.Finally, has carried on a large amount of simulation on the structure of inverse control designed in this dissertation. The first has simulated one-step non-linear system of state equation. The second has simulated the servo system by using the new structure with PID feedbacks and the open-loop structure of direct inverse control. The last, has carried on simulation of noise removal and robust. The result proved the control structure designed is rational and effectual, and has excellent robust.