Plant step data and gain estimates in the robust feedback tuning regulator problem

Estimates of the maximum gain required to retain stability of Davison's robust feedback regulator are derived based on graphical operations on plant open loop step response data. It is assumed that the exact plant model is not known, but that open loop transient step data are available. Lower bounds for the maximum gain are derived based upon the total variation of the error in modeling available plant step responses by a simple first order model. Nyquist-type criteria based on the variation of the integrated modeled error are provided by two theorems. Results indicate that the method can work well in practice, although it can be conservative in its predictions.