Closed loop neural network automatic tuner

DETAILED DESCRIPTION A closed loop autotuner is described in which the setpoint input may be changed by a user or automatically changed in order to allow the autotuner to adjust the proportional integral and/or derivative control constants of a PID controller.
In the preferred embodiment, the system employs a neural network for receiving the output of the process (or signal indicative thereof) and the current parameter settings of the PID controller.
A Smith predictor may be employed to introduce a known lag or delay into the delivery of controller output (u, to the process) if desired.
Variations may produce various combinations of proportional, integral and/or derivative parameters, singly or in combination, where such single or combination output constants are desirable.
Throughout this application, and in the claims, PID refers to a family of controllers which use some combination of PID parameters.
Different methods of generating setpoint changes for the autotuner are also described.