Nonlinear control methods supported by learning function based on neural networks

Two compensators are proposed which are effective for difficult-to-control objects having a nonlinear drift and a performance drift. For the two types of compensator a nonlinear model can be incorporated as a nonlinear action and learning functions with a neural network can be added for performance drift action. One of them is a feedforward compensator called the inverse system method, which is a method producing inverse characteristics in feedback circuits while directly introducing the nonlinear characteristics of the process into a compensator. The other is a feedback loop compensator called the K /s method. The K/s method is described. The effectiveness of the proposed method has been checked with simulation calculations