A robust QFT design for a multivariable paper machine benchmark

This paper presents a robust control system design for a nonlinear multivariable paper machine benchmark problem. The aim is to control the output variables basis weight and ash contents with the control inputs thickstock and filler flow, such that reference steps in basis weight and ash contents yield closed loop time responses within specifications for a given set of retention parameter values, with specified minimal cross influence on the other controlled variable. A static nondiagonal nonlinear precompensator is designed to minimize steady state cross coupling, then a set of equivalent linear transfer functions are identified by least squares. The time domain specifications are then translated to approximate frequency domain specifications using an assumption that the closed loop is essentially similar to a third order system. Based on the frequency domain specifications, and the set of equivalent transfer functions, a diagonal dynamic feedback controller and a diagonal dynamic prefilter are designed according to the quantitative feedback theory (QFT). This involves the computation of frequency domain constraints (Horowitz bounds) for the nominal open loops, and manual loop shaping of the latter. The main design constraint is plant delay