Simultaneous identification and control.

The established identification techniques of Astrom and Bohlin and Box and Jenkins are designed to estimate the parameters of both process and disturbance models from input/output records. The methods require open-loop experimentation on the process with a perturbation on the input from an external source. The resulting output fluctuations are often unacceptable for normal production. especially if the output exhibits drift. It is more desirable to attempt to identify and control simultaneously, avoiding the need for a prior experimental identification mode. This thesis presents some techniques of simultaneous identification and control for single-input, single-output, linear dynamic systems. The methods, based on the Box and Jenkins control scheme, are simple to apply and require a relatively small amount of computation. A particularly valuable technique exploits a separation effect for processes with large time-delays. This method of simultaneous identification and control is applied to an industrial chemical process and provides sufficient information to allow for the design of a control system which substantially reduces the output variations. Other techniques, presented in this thesis, which are not restricted by the magnitude of the process time-delay are not as yet generally applicable to industrial process-control problems.