Nonlinear model predictive control of a batch fluidized bed dryer for pharmaceutical particles

Abstract

The availability of reliable online moisture content measurements exploiting near-infrared (NIR) spectroscopy and chemometric tools allows the application of online control strategies to a wide range of drying processes in the pharmaceutical industry. In this paper, drying of particles with a pilot-scale batch fluidized bed dryer (FBD) is studied using a in-line NIR probe. A consolidated phenomenological state-space model of an FBD based on mass and energy balances is calibrated applying a nonlinear least-square identification to experimental data (grey-box modeling). Then, relying on the calibrated model, a nonlinear model predictive controller and a moving horizon state estimator are designed. The objective is to reach a specific particle moisture content setpoint at the end of the drying batch while decreasing cycle time and limiting particle temperature. A penalty term on the energy consumption can also be added to the usual tracking control cost function. Compared to a typical FBD operation in industry (mostly open-loop), it is shown that the drying time and the energy consumption can be efficiently managed on the pilot-scale process while limiting various operation problems like under drying, over drying, or particles overheating.