Twin-screw granulation uses a twin screw extruder to mix powder components and binding liquid to produce suitably sized granules that can later be compressed into tablets. The pharmaceutical industry is currently interested in this continuous process because it offers advantages over classical batch production which includes high production rates, much more consistent product quality, and reduced costs. A Quality by Design approach based on fault detection is being proposed for twin-screw granulation to create a control system that recognizes the process relationships associated with achieving quality product. The developed system in this study is founded upon fault detection and diagnosis to identify process inputs or operating parameters that cause the product to deviate from its specified quality indicators. In this study, a data-driven model of the process was used to develop an algorithm for fault detection and intervention. A reduced subset of the data was chosen to allow the reasonable assumption of local linearity around a target condition. The input variables in the model are binding liquid composition, liquid to solid ratio, solid feed rate, temperature and screw speed while the process output being used to quantify product quality is the particle size distribution of the granules. The accuracy of the fault detection and intervention algorithm is improved through simulation and experimentation in an iterative approach. The results of this study to be presented will show the potential of this method to improve process robustness and process control strategies in continuous tablet manufacture in the manner closely aligned to the FDA’s aspirations for Quality by Design.
