Investigation of Powder Blending in a Double Paddle Mixer via DEM and Experiments

<p dir="ltr">Particle mixing is a critical process in industries dealing with powders, such as pharmaceuticals and chemicals. However, understanding the complex mechanisms governing particle behavior during mixing is a significant challenge. The nonlinear nature of particle flow makes it difficult to precisely understand particle interactions. Traditional experimental methods often fall short in providing a comprehensive understanding of these intricate behaviors due to complex flow patterns and limitations in measurement techniques. To address these challenges, numerical simulations have emerged as a powerful tool, facilitated by advancements in computer technology. They offer both microscopic and macroscopic insights into particulate materials that are challenging to obtain through experiments alone. In practical industrial systems, convective mixers and twin paddle blenders are commonly utilized. Twin paddle blenders offer advantages in terms of material loading and cleaning and are suitable for a wide range of capacities. However, research on these specific mixers has been limited, and additional complexity arises from restrictions in investigating non-spherical particles and cohesive materials. Thus, this study aims to bridge this knowledge gap by investigating powder blending in double paddle mixers, utilizing a combination of Discrete Element Method (DEM) simulations and experiments.</p><p dir="ltr">The study begins by examining flow patterns and mixing mechanisms within double paddle blenders. DEM results reveal that impeller speed and initial loading patterns significantly impact mixing quality, with diffusion emerging as the primary mixing mechanism. The study delves into the role of particle characteristics, encompassing both spherical and non-spherical particles, and their influence on mixing behavior. Furthermore, it investigates the impact of mixer design parameters, such as paddle angle, width, and gap, on mixing quality, assessed through metrics like relative standard deviation (RSD) and power consumption. Moreover, the thesis addresses the calibration of DEM input parameters, with a particular focus on cohesive materials. It compares various models and identifies the Random Forest model as a means to enhance the accuracy and reliability of DEM simulations for cohesive materials. The thesis concludes by investigating the correlation between DEM results and operational parameters using machine learning techniques, providing comprehensive insight into granular mixing behaviors and their connection to operational variables.</p>