(194d) Two-Component Powder Feeding Process with Twin-Screw Feeder Can be Modeled without Information of Powder Characteristics.

In the continuous solid dosage manufacturing process, constant supply of raw materials is important to achieve a state-of-control condition. However, precise flow rate control of an active pharmaceutical ingredient (API) is difficult to control the flow rate of powder with poor flowability. To improve the flowability, it is useful to pre-mix API and excipients with good flowability before feeding. In this study, we aim to develop a feeding process model that can be used for optimizing the appropriate mixing ratio and feeder operating conditions. The target feeding process equipped with a screw feeder feeds a two-component powder mixture of an API and an excipient. We conducted feeding experiments by changing the mixing ratio of two-component powder consisting of ethenzamide (API) and spray-dried lactose (excipient), the screw rotation speed, and the powder weight in the hopper. The experimental data were used to select an appropriate model structure and explanatory variables to predict a feed factor, which is defined as the powder feed rate per screw rotation. A feed factor is helpful to determine the appropriate operating conditions and to control the flow rate. The results showed that the gray-box model, which combines a physical model and a statistical model that expresses the parameters of the physical model, had a smaller prediction error than the other models such as partial least squared regression models, random forest models, and gaussian process regression models. The results also showed that the prediction error of the models with the screw rotation speed, the powder weight in the hopper, and the mixing ratio as explanatory variables was almost the same as that of the models with powder properties as explanatory variables in addition to these variables.