A hybrid system for design space estimation in a rotary tablet press

The design space (DS) identification in a pharmaceutical process is a crucial aspect of the quality-by-design (QbD) initiative. For computationally inexpensive models, the original process model can be used to identify the DS. However, a hybrid approach is useful when the original model is insufficient for a comprehensive estimate of the design space. The Tablet Press (TP) is a critical unit operation that directly compresses powder blends into final product units and allows the study of the impact of different additives, such as glidants or lubricants, on tablet quality. This work presents a hybrid system to identify the design space of process conditions and material attributes in a TP while ensuring that the desired tablet quality constraints are met. A semi-mechanistic model is efficient for glidant-based blends to characterize the design space. However, for lubricated blends, it is found that a data-driven response surface model allows significantly better identification of the design space to produce quality tablets. The resultant design space contours are based on different quality levels, which include model uncertainty, and material-process conditions interaction with product quality attributes that help reduce the high cost of comprehensive DS experiments. Finally, the estimated design spaces are experimentally studied, showing the hybrid system's efficacy relative to a mechanistic approach.