| Food extrusion is a multi-input multi-output (MIMO) process characterized by time-variant dynamics, long and variable time delays, nonlinearity, and input constraints. In this research, multivariable generalized predictive control (MGPC) was developed and applied in food extrusion control. First, a technique named as the max-max scheme was developed to deal with problems and difficulties resulting from unequal time in multivariable predictive control. Control accuracy and stability are improved by the scheme. Then, a generalized dual-target predictive control algorithm was developed to allow specification and tracking of both input and output set points in an optimal manner while limiting the size of input moves. Along with on-line recursive least-squares system identification, these techniques were implemented in control of a twin-screw food extruder. Experiment results prove the effectiveness of the techniques developed. |
