A model-free adaptive controller is disclosed which uses a dynamic artificialneural network with a learning algorithm (22) to controlany single-variable or multivariable open-loop stable, controllable, andconsistently direct-acting or reverse-acting industrial process (76)without requiring any manual tuning, quantitative knowledge of the process, orprocess identifiers. The need for process knowledge isavoided by substituting "1" for the actual sensitivity function ~y(t)/~u(t) ofthe process.
展开▼
