Abstract
Although neural network controllers have high performance of controllability, it is difficult to guarantee stability due to their nonlinearity and complexity. We showed a design method of structures of neural network controllers for active control of vibration systems considering not only controllability but also stability. The structure of neural networks was layered and designed to identify linear single-degree-of-freedom (SDOF) systems for vibration control on the basis of the equation form of the systems with linear acceleration method which is one of numerical integration methods. The performance of the controller was compared with that of a turned PID controller.
