文摘
Most milling circuits can be well approximated using linear process models, but the mill itself tends to exhibit strong non-linear behavior. To overcome this, each section of the milling circuit is typically controlled using a linear controller and the non-linearities are handled by non-linear capable supervisory layer, such as fuzzy logic or expert rules. Mintek developed the StarCS RNMPC that can use both linear and non-linear models to enable the entire milling circuit to be handled by one controller. MPC is inherently optimising and can therefore provide stabilisation as well as optimisation of the milling circuit, while adhering to all process constraints. This paper shows a simulation of a milling circuit that is stabilised and optimised using a single StarCS RNMPC controller with linear and non-linear prediction models as well as other unique features, such as cascaded CV-to-CV prediction models.