An intelligent controller using a self-learning neural net

This paper describes a self-learning neural controller that learns the behavior of a system or plant and makes decisions to reach the desired state. The components of the controller are a data acquisition system with a knowledge base, a control net and a scheduler. Self-learning of the controller can be in two modes: off-line and on-line. In both modes, the data acquisition system stores the learnt data in the knowledge base in the form of system state vectors. The control net is trained with learnt patterns decided by the scheduler. After training, the control net with the current weight vectors is switched on for operation