(509b) Hybrid Theoretical-Empirical Approach to Modelling of Twin Screw Feeders for Continuous Tablet Manufacturing

Screw feeders are popular equipment for the handling and metering of bulk solids in the pharmaceutical industry, especially in continuous tablet manufacturing. Predictive models are crucial to meet quality requirements, however a limited number of predictive process models have been developed so far in the scientific literature. Despite the increasing research in the last decade in the manufacturing of powder-based products, there is still a lack of knowledge on the physics governing the dynamic behaviour of the screw feeders for different material properties, screw designs and operating conditions. As a result, first-principles models for a complex system such as powder feeding have not been properly validated in both open and closed loop operations. Surrogate models, multivariate techniques and empirical correlations have often been used instead to tackle process design, optimisation and control applications.

In this work, a mathematical modelling of twin screw feeders, suitable for continuous drug manufacturing processes, has been developed using a combined physics-based- and data-driven approach. The aim of the model is to predict how the screw feeder design, the hopper fill level, the screw speed and the physical properties affect the dynamic behaviour of the open loop system. A First Order Plus Dead Time (FOPDT) model has been developed in MATLAB (MathWorks) where the vertical stress exerted by the powder in the hopper, estimated from the equilibrium of forces on slice elements of infinitesimal thickness, has been assumed to affect the powder densification within the screw feeder and, thus, the mass flow rate provided by the feeder. A method to predict the amplitude of the stochastic fluctuations in the mass flow rate observed in the process data has been suggested and included in the model. Model parameters have been estimated for a total of six different bulk solids (two non-cohesive powders, two cohesive ones and two in the transitional range, categorised according to their Hausner ratio values). Furthermore, two different screw sizes and two hopper geometries have been investigated. The model predictions are in good agreement with the experimental data sets. The model can be explored to investigate the mass flow – screw feed relationships and the optimal range of operating conditions. It also may be used for online control, to identify the original equipment manufacturer control system and to design alternative control strategies. Furthermore, the model can be exploited to optimise refill operations, when the process is not controlled and the system is running under volumetric mode (i.e. open loop, the screw speed is constant and the controller is temporarily off, as the weight in hopper is increasing).