As various industries continue to develop complex, fundamental process mode
ls, there exists a need to systematically incorporate these complex models
into the controller design. Three model predictive controllers (MPC), each
incorporating internal models with varying degrees of complexity, is applie
d to a nonlinear, fundamental, continuous pulp digester "plant." The first
two controllers utilize linear models, one obtained through subspace identi
fication and the other obtained from the linearization of the fundamental m
odel. The third model predictive controller uses the complex, nonlinear dig
ester model with extended linearization to update the controller model for
future predictions and control computations. The two MPC controllers based
on the fundamental model, both linear and nonlinear, had better closed-loop
performance than the controller utilizing the subspace identified model. T
he closed-loop performance of the linear and nonlinear MPC controllers (bas
ed on the fundamental model) were indistinguishable for stochastic disturba
nce rejection.