Analysis of particle motion and mixing in the powder chamber of a capsule filling machine with different stirrer designs using DEM simulations

Abstract

The stirrer geometry has been experimentally shown to influence remarkably powder behaviour in the chamber of a capsule filling machine with a vacuum drum. This study investigates this influence using Discrete Element Method (DEM) simulations. First, DEM parameters of a lactose grade were calibrated via an auger dosing system, using a design of experiments approach and optimization to identify parameter combinations that replicate experimental bulk behaviour. The powder chamber with different stirrers was then numerically analysed. Experiments were performed placing a tracer in the middle of the chamber, and after one minute of stirrer rotation, the axial distribution of the tracer was measured and compared with simulations using the calibrated powder. While some differences were observed between the experimental and simulation results, reasonable agreement was achieved. Despite the discrepancies, the current calibration is sufficient for a comparative analysis of stirrer geometries and particle movement within the powder chamber. The Lacey mixing index, total and net displacement of the particles, average particle velocities, and velocity fields were considered. Simulations provided detailed insights at the particle level, which would be challenging to obtain experimentally, revealing differences in particle behaviour depending on stirrer geometry. The results obtained can help to further optimize stirrer geometries in the future.

Introduction

Solid dosage formulations are used for administering drug substances, particularly via the oral route, which is considered the most patient-compliant and conventional method [1,2]. In addition, for pulmonary drug delivery, the active pharmaceutical ingredient (API) is filled into capsules, which are then placed into a dry powder inhaler (DPI) [3]. Filling powder into hard capsules has several advantages, such as enhanced bioavailability due to increased porosity, less demanding powder flow requirements compared to tableting, and the ability to fill formulations with lower compressibility [4]. Capsule filling can be classified as manually, semi-automatic and fully automated [4,5] and can be categorized based on how the material dose is measured. Capsule-dependent machines measure the dose based on the capsule size, while capsule-independent machines measure the dose separately [2]. Common capsule filling principles are the auger filling, vibratory filling, dosator nozzle, dosing disc and vacuum drum filling [2,4,6].
The vacuum drum filler is a system commonly used for micro-dosing due to its capability for precise dosing down to 1 mg. This system is particularly prevalent in dry powder inhalation (DPI). The vacuum drum filling principle applies vacuum to fill powder into the bores of a drum, which then rotates, and the powder is discharged into the capsules below. A machine using this filling method mainly consists of a powder chamber with a stirrer positioned above the vacuum drum (see Fig. 1). Previous experimental investigations have primarily focused on analysing the influence of different powder properties on capsule filling weight and the resulting relative standard deviation (RSD), as well as the impact of process parameters on these outcomes [7,8].

In our previous study [9], the mixing behaviour in the powder chamber of the capsule filling machine was investigated to understand how the powder bed distributes along the chamber’s length. A screw feeding system at the top continuously feeds powder into the central section of the chamber. The stirrer must induce axial mixing to maintain a homogeneous powder bed consistency and height, thereby preventing accumulation in specific regions, such as the middle section. To replicate the actual conditions under which the powder is introduced, a tracer powder was placed in the middle of the chamber. After a defined mixing time, samples were taken along the chamber’s length to analyse the uniformity of the axial distribution. The small size of the chamber, along with the complex geometry of the stirrer, which includes its core, wires, and spikes, makes radial sampling particularly challenging. Accessing different radial positions would be difficult and could compromise the integrity of the sample collection process. This, however, is an issue that would not arise in simulations, where more information can be gathered.

The Discrete Element Method (DEM) offers the possibility to perform the aforementioned simulations and has become widely used, particularly in chemical engineering [10], as well as in pharmaceutical applications, such as die filling prior to tablet compaction [11] and capsule filling with dosator [12,13]. To the best of our knowledge, no studies have explored DEM simulations of capsule filling using a vacuum drum. Existing research focuses on other methods, such as the work by Loidolt et al. [12], who used DEM simulations to study dosator nozzle filling. In this method, the nozzle dips into a powder bed until it reaches a specified distance from the bottom, allowing powder to enter and be compressed to a predefined degree. The filled nozzle is then lifted and transfers the powder into an open capsule. Loidolt et al. [12] found that powder properties, including particle size and flow behaviour, significantly affect the dosed mass. However, the authors did not validate their simulations against experimental data. Building on this work, Madlmeir et al. [13] investigated the critical process parameters of a dosator system for capsule filling using DEM simulations. They calibrated and validated their models, finding that while the models accurately predict capsule filling for the tested powder, the powder bed significantly impacts filling weight. The research further suggests that although the dosator process is primarily volume-based, factors such as dosator geometry and proportion also influence capsule filling weight.

The focus of our research is on the powder chamber with a stirrer on top of the vacuum drum rather than on the capsule filling itself. Although no studies were found on DEM for this specific setup, several studies have analysed particle motion in different types of mixers. Hassanpour et al. [14] studied particle flow in a paddle mixer by comparing DEM simulations with experimental data using Positron Emission Particle Tracking (PEPT). They found good qualitative agreement between the two methods in terms of flow patterns, although DEM produced smoother velocity distributions, while PEPT data was more scattered due to the single particle tracking method. Zhang et al. [15]

examined particle movements in a flighted rotating drum (FRD) using Particle Tracking Velocimetry (PTV) and DEM simulations. They varied the drum’s rotating speed and filling degree, analysing how these factors affected the falling time distribution of particles. Ebrahimi et al. [16] used DEM to study the impact of different impeller configurations on mixing performance in a horizontal paddle mixer. They found that the mixing performance was significantly influenced by the angle of the paddles. Jadidi et al. [17] explored how paddle angle, width, and gap affect the mixing performance of a twin paddle blender using DEM. They evaluated several parameters, including mixing index, velocity field, diffusivity, granular temperature, force on particles, and power consumption. Boonkanokwong et al. [18] investigated the effect of the number of impeller blades on granular flow in a bladed mixer. They found that the number of blades significantly impacts the granular flow and mixing kinetics, as well as various velocity and contact force distributions.

In this study, DEM simulations of the powder chamber of a capsule filling machine with a vacuum drum were conducted using various stirrer geometries. The objective was to gain a better understanding of particle behaviour and the impact of stirrer design by analysing the mixing index, velocity fields, and particle motion. To achieve this, the DEM was calibrated for a lactose powder, and performed simulations were validated against experimental results. It is important to note that many simplifications were necessary to set up simulations that could be run within a reasonable computational time. Consequently, the modelling of mixing in the powder chamber represents only the initial step toward the design and optimization of stirrers and the capsule filling process using a vacuum drum.

This article begins by explaining the numerical method employed for the DEM simulations, including the relevant equations and contact model. Next, the calibration process for the input parameters is detailed, outlining the experiments and simulations conducted as well as the corresponding calibration results. Subsequently, the study of the powder chamber is introduced, describing the experimental setup with tracers used for simulation validation. Finally, the analysis focuses on mixing, particle behaviour, and velocity fields with different stirrer geometries, offering insights into the effects of stirrer design.