All4Nutra
Abstract
An industrial-scale pharmaceutical powder blending process was studied via discrete element method (DEM) simulations. A DEM model of two active pharmaceutical ingredient (API) components and a combined excipient component was calibrated by matching the simulated response in a dynamic angle of repose tester to the experimentally observed response. A simulation of the 25-minute bin blending process predicted inhomogeneous API distributions along the rotation axis of the blending container. These concentration differences were confirmed experimentally in a production-scale mixing trial using high-performance liquid chromatography analysis of samples from various locations in the bin. Several strategies to improve the blend homogeneity were then studied using DEM simulations. Reversing the direction of rotation of the blender every minute was found to negligibly improve the blending performance. Introducing a baffle into the lid at a 45° angle to the rotation axis sped up the axial mixing and resulted in a better final blend uniformity. Alternatively, rotating the blending container 90° around the vertical axis five minutes prior to the process end was predicted to reduce axial segregation tendencies.
Introduction
Powder mixing is a crucial unit operation in a variety of industries, such as pharmaceuticals, chemicals, plastics and food. In the pharmaceutical industry, blending in intermediate bulk containers (IBCs) is used extensively in oral dosage form manufacturing. The IBCs are rotated to blend the powder and are typically shaped as a cylinder with a capped cone or a cuboid with a capped pyramid, also referred to as a tote blender. They come in a wide range of sizes, from a few liters on the laboratory scale to several cubic meters on commercial scales. The key concerns in powder mixing processes are the mixing time and the blend uniformity (Suzanne et al., 2016).
Despite their prevalence, these types of blending processes are not understood sufficiently well to be designed or scaled up reliably. The blender’s geometry and operating conditions such as the rotation speed, fill level, loading strategy and mixing time are crucial in determining the blending performance. Newly designed processes or processes that have been transferred to a different scale often fall short of meeting the mixing quality requirements. Poor understanding of the interplay between the properties of individual powder components, the geometry and the process conditions can lead to expensive and time-consuming trials to establish a robust mixing process.
Experimental studies, in particular studies on the laboratory scale, have been employed to better understand powder mixing. The blend homogeneity in these studies is commonly evaluated via analytical methods, such as high-performance liquid chromatography (HPLC), with samples taken from the process using a sample thief. This approach is simple yet cumbersome and leads to a low spatial and temporal resolution. In addition, due to the invasiveness of the method, the powder bed is disturbed and the blending process is altered. Non-invasive approaches (e.g. using near-infrared spectroscopy) to measure the blend homogeneity have been developed, but they are not commonly implemented in pharmaceutical batch blending processes (Nadeem and Heindel, 2018). Moreover, they can only provide information about the blending state of the part of the powder bed that is exposed to the sensor, which may lead to misguided assumptions about the blending state of the entire powder bed (Scheibelhofer, 2010).
Over the last decades, this lack of spatially and temporally resolved blend uniformity data has motivated a multitude of computational studies of powder blending processes to provide process insights (Jadidi et al., 2022, Jin and Shen, 2024). Granular blending processes are typically studied using the discrete element method (DEM), which describes the behavior of a powder system by integrating Newton’s laws of motion for individual particles (Cundall and Strack, 1979). DEM simulations can be a valuable tool for the design and scale-up of powder blending since they allow for fast screening of the process conditions and their influence on the mixing process. The local mixing quality can be computed at any time and any point in the device (Lemieux et al., 2007, Arratia et al., 2006).
DEM simulations of industrial powder processes that take into account every single particle are prohibitively expensive and slow. To decrease the computational requirements, particles of industrially relevant powders are usually scaled up in the simulation and the size distribution is kept narrow (Carr et al., 2023). The particles are typically assumed to be spherical to simplify contact detection. For most industrially relevant granular materials, the stiffness in the simulations is orders of magnitude lower than the physical material stiffness to avoid unfeasibly small simulation time steps (Ketterhagen and Wassgren, 2022, Coetzee, 2019).
Due to these assumptions regarding shape, size and stiffness of the simulated particles, using direct measurements of the physical properties of pharmaceutical powders as DEM parameters is unreasonable. Instead, the macroscopic powder behavior is measured in small-scale tests and the DEM parameters are tuned to reproduce the observed bulk powder response (Ketterhagen and Wassgren, 2022, Coetzee, 2016). Depending on the powder properties, application of interest and available equipment, a variety of bulk powder tests has been used to calibrate DEM parameters for granular materials: static and dynamic angle of repose tests, uniaxial compression tests, shear cells, and others (Ketterhagen and Wassgren, 2022). Until recently, DEM calibration routines for bulk powders have predominantly followed iterative approaches (Orefice and Khinast, 2020, Coetzee, 2017). Recent work has based the calibration on databases of DEM parameter sets and their modeled response in bulk powder tests. Regression models have been used to correlate the bulk powder response to the DEM parameters (Benvenuti et al., 2016, Forgber et al., 2022).
In the last twenty years, DEM codes that rely on graphics processing hardware have led to a significant performance increase (Radeke et al., 2010, Gan et al., 2016). Most numerical work on industrial powder blending employed DEM simulations to describe the entire duration of the industrial process (Sen et al., 2017, Yu et al., 2021). However, industrial-length blending processes have occasionally been modeled by simulating a short initial process time using the DEM and extrapolating from the obtained powder mechanics (Siegmann et al., 2021, Bednarek et al., 2019, Mostafaei et al., 2023). The simulations provide detailed information about the dynamics and the local composition of the powder blend in the system.
A thorough review of existing DEM studies of granular blending has been published by Jadidi et al (Jadidi et al., 2022), and a general review of computational and experimental studies on granular mixing is available by Jin and Shen (Jin and Shen, 2024). Most DEM studies on powder blending in bin blenders have focused on free-flowing granular materials. High fill levels have been found to be detrimental to the mixing performance (Lemieux et al., 2007, Arratia et al., 2006, Ren et al., 2013). The effect of the rotation rate on the mixing of free-flowing granular material has been studied in various works, and a low mixing performance at very low and very high rotation rates has been suggested (Ren et al., 2013). Similar observations about the influence of fill level and rotation rate have been made based on DEM simulations of V-blenders (Tahvildarian et al., 2013). A contrast between fast convection in the radial direction and slow dispersion in the direction of the rotation axis has been described (Arratia et al., 2006, Alizadeh et al., 2014). The introduction of baffles into tote blenders has been shown to increase axial mixing in DEM simulations (Yu et al., 2021), but an experimental study has raised concerns that the baffles might hinder the mixing performance depending on the geometry, material, and operating conditions (Arratia et al., 2006). Only few DEM studies have considered the blending of cohesive material in bin blenders (Mostafaei et al., 2023, Tanabe et al., 2019) and rotating drums (Govender et al., 2023, Yazdani and Hashemabadi, 2019). These studies have modeled short process times to gain an understanding of the mixing dynamics or to extrapolate the blending performance of an industrially relevant process duration.
This study focuses on a production-scale batch blending process in a pharmaceutical direct compaction line. The DEM was used to model the blending of a ternary mixture of cohesive granular materials in a large tote blender over the full industrial process length. The blend homogeneity is paramount for being able to compress the powder into stable tablets and keeping the active pharmaceutical ingredient (API) dosage in those tablets within the specification (Radl et al., 2010). The distribution of two APIs and a homogeneous excipient component in the IBC over time was studied in the simulations. The calibration of the simulation model followed a calibration protocol based on a database of DEM calibration tests as suggested previously (Benvenuti et al., 2016, Forgber et al., 2022). The predicted mixing trends were compared to HPLC assays of the APIs collected during a production-scale mixing trial at different time points and different locations in the blending bin. The insight gained from the simulations was used to adjust the sampling process. Additional DEM simulations were performed to provide guidance on how to improve the performance and robustness of the blending process.
Read more here
Benedict Benque, Luca Orefice, Thomas Forgber, Matthias Habeler, Beate Schmid, Johan Remmelgas, Johannes Khinast, Improvement of a pharmaceutical powder mixing process in a tote blender via DEM simulations,
International Journal of Pharmaceutics, Volume 658, 2024, 124224, ISSN 0378-5173, https://doi.org/10.1016/j.ijpharm.2024.124224.
Read more on Pharmaceutical Powder Mixing Process here and also the following article:
Effects of processing parameters and blade patterns on continuous pharmaceutical powder mixing
